Multi_Layer_Perceptrons.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Multi Layer Perceptrons.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/dipanjannC/Deep-Learning/blob/master/Multi_Layer_Perceptrons.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A6cVT22pWDNs",
        "colab_type": "text"
      },
      "source": [
        "#Initializations"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tgzoh1VmM5Ag",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from tensorflow.examples.tutorials.mnist import input_data\n",
        "import tensorflow as tf"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VoMbc5OHW2xU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 570
        },
        "outputId": "b64a07a0-5b61-47a6-b691-d37f49ca4ab1"
      },
      "source": [
        "mnist = input_data.read_data_sets(\"MNIST_data/\",one_hot=True)"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From <ipython-input-4-758d29429358>:1: read_data_sets (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:260: maybe_download (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please write your own downloading logic.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/base.py:252: _internal_retry.<locals>.wrap.<locals>.wrapped_fn (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use urllib or similar directly.\n",
            "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:262: extract_images (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use tf.data to implement this functionality.\n",
            "Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
            "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:267: extract_labels (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use tf.data to implement this functionality.\n",
            "Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:110: dense_to_one_hot (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use tf.one_hot on tensors.\n",
            "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
            "Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
            "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
            "Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:290: DataSet.__init__ (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9lVmSE75Xfbd",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# %matplotlib notebook\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import time\n",
        "# https://gist.github.com/greydanus/f6eee59eaf1d90fcb3b534a25362cea4\n",
        "# https://stackoverflow.com/a/14434334\n",
        "# this function is used to update the plots for each epoch and error\n",
        "def plt_dynamic(x, y, y_1, ax, ticks,title, colors=['b']):\n",
        "    ax.plot(x, y, 'b', label=\"Train Loss\")\n",
        "    ax.plot(x, y_1, 'r', label=\"Test Loss\")\n",
        "    if len(x)==1:\n",
        "        plt.legend()\n",
        "        plt.title(title)\n",
        "    plt.yticks(ticks)\n",
        "    fig.canvas.draw()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1-hT4St_beLh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        },
        "outputId": "a3285277-f175-4c31-88cf-a6c4dba4c361"
      },
      "source": [
        "print(\"number of datapoints: \", mnist.train.images.shape[0],\" number of pixels in each image: \",mnist.train.images.shape[1])\n",
        "print(\"number of datapoints: \", mnist.train.labels.shape[0],\" number of labels in dataset: \",mnist.train.labels.shape[1])"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "number of datapoints:  55000  number of pixels in each image:  784\n",
            "number of datapoints:  55000  number of labels in dataset:  10\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fLMKvHhXX8pZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Network Parameters\n",
        "n_input = 784\n",
        "n_classes = 10\n",
        "n_hidden_1 = 512 # number of neurons in 1st Layer\n",
        "n_hidden_2 = 128 # number of neurons in 2nd layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ynFshda3c2hu",
        "colab_type": "text"
      },
      "source": [
        "So , our architecture is iike,\n",
        "\n",
        "    F.C        F.C      SMC(softmax classifier)\n",
        "728  ---------->  512  ----------->  128 ----------->   10"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7aXPoSFQc1en",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Initializing the Placeholders\n",
        "x = tf.placeholder(tf.float32,[None, 784])  # placeholder for inputs\n",
        "y_ = tf.placeholder(tf.float32,[None, 10])  # placeholder for actual y_i\n",
        "\n",
        "#placeholder for dropouts\n",
        "keep_prob = tf.placeholder(tf.float32)\n",
        "keep_prob_input = tf.placeholder(tf.float32)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ad5VoBhbcn7b",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 92
        },
        "outputId": "b231f905-52ba-4b18-d035-1a0b6fe4bb73"
      },
      "source": [
        "# Weight initialization\n",
        "\n",
        "# https://arxiv.org/pdf/1707.09725.pdf#page=95\n",
        "# https://www.tensorflow.org/api_docs/python/tf/random_normal\n",
        "# Outputs random values from a normal distribution mean=0 std=1\n",
        "\n",
        "# If we sample weights from a normal distribution N(0,σ) we satisfy this condition with σ=√(2/(ni+ni+1). \n",
        "# h1 =>  σ=√(2/(fan_in+fan_out+1) = 0.039  => N(0,σ) = N(0,0.039)\n",
        "# h2 =>  σ=√(2/(fan_in+fan_out+1) = 0.055  => N(0,σ) = N(0,0.055)\n",
        "# out =>  σ=√(2/(fan_in+fan_out+1) = 0.120  => N(0,σ) = N(0,0.120)\n",
        "# SGD: Xavier/Glorot Normal initialization.\n",
        "weights_sgd = {\n",
        "    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1],stddev=0.039, mean=0)),    #784x512 # sqrt(2/(784+512)) = 0.039\n",
        "    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2],stddev=0.055, mean=0)), #512x128 # sqrt(2/(512+128)) = 0.055\n",
        "    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes],stddev=0.120, mean=0))  #128x10\n",
        "}\n",
        "\n",
        "# https://arxiv.org/pdf/1707.09725.pdf#page=95\n",
        "# for relu lates\n",
        "# If we sample weights from a normal distribution N(0,σ) we satisfy this condition with σ=√(2/(ni). \n",
        "# h1 =>  σ=√(2/(fan_in+1) = 0.062  => N(0,σ) = N(0,0.062)\n",
        "# h2 =>  σ=√(2/(fan_in+1) = 0.125  => N(0,σ) = N(0,0.125)\n",
        "# out =>  σ=√(2/(fan_in+1) = 0.120  => N(0,σ) = N(0,0.120)\n",
        "# He Normal initialization.\n",
        "weights_relu = {\n",
        "    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1],stddev=0.062, mean=0)),    #784x512\n",
        "    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2],stddev=0.125, mean=0)), #512x128\n",
        "    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes],stddev=0.120, mean=0))  #128x10\n",
        "}\n",
        "biases = {\n",
        "    'b1': tf.Variable(tf.random_normal([n_hidden_1])),             #512x1\n",
        "    'b2': tf.Variable(tf.random_normal([n_hidden_2])),             #128x1\n",
        "    'out': tf.Variable(tf.random_normal([n_classes]))              #10x1\n",
        "}"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TMdyhZXLj0zL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Parameters\n",
        "training_epochs = 50\n",
        "learning_rate = 0.001\n",
        "batch_size = 100\n",
        "display_step = 1 #i.e after how many steps you want to display your plot,here after each epoch we will display \n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JxygHn2iqpNN",
        "colab_type": "text"
      },
      "source": [
        "# Model 1 : input (784) - sigmoid(512) - sigmoid(128) - softmax(output 10) "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CnLXKquKqBhO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Create model\n",
        "def multilayer_perceptron(x, weights, biases):\n",
        "    # Use tf.matmul instead of \"*\" because tf.matmul can change it's dimensions on the fly (broadcast)\n",
        "    print( 'x:', x.get_shape(), 'W[h1]:', weights['h1'].get_shape(), 'b[h1]:', biases['b1'].get_shape())        \n",
        "    \n",
        "    # Hidden layer with Sigmoid activation\n",
        "    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) #(x*weights['h1']) + biases['b1']\n",
        "    layer_1 = tf.nn.sigmoid(layer_1)\n",
        "    print( 'layer_1:', layer_1.get_shape(), 'W[h2]:', weights['h2'].get_shape(), 'b[h2]:', biases['b2'].get_shape())        \n",
        "    \n",
        "    # Hidden layer with Sigmoid activation\n",
        "    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) # (layer_1 * weights['h2']) + biases['b2'] \n",
        "    layer_2 = tf.nn.sigmoid(layer_2)\n",
        "    print( 'layer_2:', layer_2.get_shape(), 'W[out]:', weights['out'].get_shape(), 'b3:', biases['out'].get_shape())        \n",
        "    \n",
        "    # Output layer with Sigmoid activation\n",
        "    out_layer = tf.matmul(layer_2, weights['out']) + biases['out'] # (layer_2 * weights['out']) + biases['out']    \n",
        "    out_layer = tf.nn.sigmoid(out_layer)\n",
        "    print('out_layer:',out_layer.get_shape())\n",
        "\n",
        "    return out_layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ucWR-xHl3n7g",
        "colab_type": "text"
      },
      "source": [
        "Model 1 with Adam Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ev19wDrpwdXi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2224
        },
        "outputId": "1f1f5ed5-2113-409a-db55-eae80795391b"
      },
      "source": [
        "# Since we are using sigmoid activations in hiden layers we will be using weights that are initalized as weights_sgd\n",
        "# Computer y_i hat i.e the predicted value\n",
        "y_sgd = multilayer_perceptron(x,weights_sgd,biases)\n",
        "\n",
        "#calculating the log loss value i.e cross entropy\n",
        "cost_sgd = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = y_sgd, labels = y_))\n",
        "\n",
        "# Initializting the Optimizers\n",
        "optimizer_adam = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost_sgd)\n",
        "optimizer_sgdc = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost_sgd)\n",
        "\n",
        "\n",
        "with tf.Session() as sess:\n",
        "    # initializing all the global variables\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.   \n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size) # fetching the no. of batches according to batch_size we initialized earlier\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here c: corresponds to the parameter cost_sgd\n",
        "            # w : correspondse to the parameter weights_sgd\n",
        "            # c = sess.run() return the cost after every bath during train\n",
        "            # w = sess.run() return the weights that are modified after every batch through Back prop\n",
        "            # w is dict w = {'h1': updated h1 weight vector after the current batch,\n",
        "            #                'h2': updated h2 weight vector after the current batch, \n",
        "            #                'out': updated output weight vector after the current batch, \n",
        "            #                }\n",
        "            # you check these w matrix for every iteration, and check whats happening during back prop\n",
        "            #\n",
        "            # note: sess.run() returns parameter values based on the input parameters\n",
        "            # _, c, w = sess.run([optimizer_adam, cost_sgd,weights_sgd]) it returns three parameters\n",
        "            # _, c = sess.run([optimizer_adam, cost_sgd ]) it returns two parameters\n",
        "            # _ = sess.run([optimizer_adam]) it returns one paramter (for the input optimizer it return none)\n",
        "            # c = sess.run([cost_sgd]) it returns one paramter (for the input cost return error after the batch)\n",
        "\n",
        "            # feed_dict={x: batch_xs, y_: batch_ys} here x, y_ should be placeholders\n",
        "            # x, y_ are the input parameters on which the models gets trained\n",
        "\n",
        "            # here we use AdamOptimizer\n",
        "            _, c, w = sess.run([optimizer_adam, cost_sgd,weights_sgd], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch # calulating average training loss after each iteration(epoch)\n",
        "            c = sess.run(cost_sgd, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch # calulating average training loss after each iteration(epoch)\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-sigmoid(512)-sigmoid(128)-sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-sigmoid(512)-sigmoid(128)-sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(y_sgd,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))\n",
        "\n"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x: (?, 784) W[h1]: (784, 512) b[h1]: (512,)\n",
            "layer_1: (?, 512) W[h2]: (512, 128) b[h2]: (128,)\n",
            "layer_2: (?, 128) W[out]: (128, 10) b3: (10,)\n",
            "out_layer: (?, 10)\n",
            "Epoch: 0001 train cost=1.682941594 test cost=1.676739588\n",
            "Epoch: 0002 train cost=1.538383644 test cost=1.536499198\n",
            "Epoch: 0003 train cost=1.516285429 test cost=1.516714223\n",
            "Epoch: 0004 train cost=1.504306501 test cost=1.505756942\n",
            "Epoch: 0005 train cost=1.495986246 test cost=1.498500832\n",
            "Epoch: 0006 train cost=1.489618614 test cost=1.493705590\n",
            "Epoch: 0007 train cost=1.484975125 test cost=1.490166647\n",
            "Epoch: 0008 train cost=1.481422061 test cost=1.487790055\n",
            "Epoch: 0009 train cost=1.478575771 test cost=1.485733355\n",
            "Epoch: 0010 train cost=1.476058757 test cost=1.484414918\n",
            "Epoch: 0011 train cost=1.474140810 test cost=1.483432052\n",
            "Epoch: 0012 train cost=1.472469109 test cost=1.482476808\n",
            "Epoch: 0013 train cost=1.471253896 test cost=1.482072009\n",
            "Epoch: 0014 train cost=1.470228230 test cost=1.481680430\n",
            "Epoch: 0015 train cost=1.469267078 test cost=1.481633327\n",
            "Epoch: 0016 train cost=1.468294699 test cost=1.481299715\n",
            "Epoch: 0017 train cost=1.467557364 test cost=1.480899447\n",
            "Epoch: 0018 train cost=1.466911943 test cost=1.480811645\n",
            "Epoch: 0019 train cost=1.466266738 test cost=1.480263410\n",
            "Epoch: 0020 train cost=1.466055585 test cost=1.480381988\n",
            "Epoch: 0021 train cost=1.465652669 test cost=1.480519126\n",
            "Epoch: 0022 train cost=1.465357198 test cost=1.480613761\n",
            "Epoch: 0023 train cost=1.465114465 test cost=1.480731448\n",
            "Epoch: 0024 train cost=1.464870861 test cost=1.480987668\n",
            "Epoch: 0025 train cost=1.465327208 test cost=1.480789606\n",
            "Epoch: 0026 train cost=1.464503691 test cost=1.480841666\n",
            "Epoch: 0027 train cost=1.464421657 test cost=1.480871884\n",
            "Epoch: 0028 train cost=1.464904290 test cost=1.480552015\n",
            "Epoch: 0029 train cost=1.464153756 test cost=1.480324412\n",
            "Epoch: 0030 train cost=1.464373003 test cost=1.480429038\n",
            "Epoch: 0031 train cost=1.463770767 test cost=1.480009640\n",
            "Epoch: 0032 train cost=1.463773603 test cost=1.480380417\n",
            "Epoch: 0033 train cost=1.464417555 test cost=1.480926221\n",
            "Epoch: 0034 train cost=1.463737363 test cost=1.480057773\n",
            "Epoch: 0035 train cost=1.464048922 test cost=1.480330759\n",
            "Epoch: 0036 train cost=1.463738313 test cost=1.480122780\n",
            "Epoch: 0037 train cost=1.463548761 test cost=1.479790859\n",
            "Epoch: 0038 train cost=1.463763268 test cost=1.479964185\n",
            "Epoch: 0039 train cost=1.463622603 test cost=1.479820975\n",
            "Epoch: 0040 train cost=1.463226316 test cost=1.479336784\n",
            "Epoch: 0041 train cost=1.463456617 test cost=1.479563129\n",
            "Epoch: 0042 train cost=1.463210438 test cost=1.478993087\n",
            "Epoch: 0043 train cost=1.463667696 test cost=1.479659577\n",
            "Epoch: 0044 train cost=1.463580203 test cost=1.479884903\n",
            "Epoch: 0045 train cost=1.463130680 test cost=1.479209472\n",
            "Epoch: 0046 train cost=1.463343211 test cost=1.479679087\n",
            "Epoch: 0047 train cost=1.463521476 test cost=1.479831500\n",
            "Epoch: 0048 train cost=1.463248908 test cost=1.478998889\n",
            "Epoch: 0049 train cost=1.463021111 test cost=1.478639287\n",
            "Epoch: 0050 train cost=1.463166281 test cost=1.479057719\n",
            "Epoch: 0051 train cost=1.462935871 test cost=1.478824414\n",
            "Epoch: 0052 train cost=1.463258927 test cost=1.480008315\n",
            "Epoch: 0053 train cost=1.463251598 test cost=1.479993675\n",
            "Epoch: 0054 train cost=1.462938199 test cost=1.479079332\n",
            "Epoch: 0055 train cost=1.462858350 test cost=1.479308928\n",
            "Epoch: 0056 train cost=1.462809785 test cost=1.479209700\n",
            "Epoch: 0057 train cost=1.462790944 test cost=1.479166078\n",
            "Epoch: 0058 train cost=1.462787897 test cost=1.479093847\n",
            "Epoch: 0059 train cost=1.463507203 test cost=1.479757074\n",
            "Epoch: 0060 train cost=1.463396820 test cost=1.479302944\n",
            "Epoch: 0061 train cost=1.462859743 test cost=1.479120564\n",
            "Epoch: 0062 train cost=1.462731714 test cost=1.478972243\n",
            "Epoch: 0063 train cost=1.462709075 test cost=1.478847529\n",
            "Epoch: 0064 train cost=1.462706510 test cost=1.478790617\n",
            "Epoch: 0065 train cost=1.463417166 test cost=1.479749906\n",
            "Epoch: 0066 train cost=1.463001789 test cost=1.479726478\n",
            "Epoch: 0067 train cost=1.462877967 test cost=1.479089723\n",
            "Epoch: 0068 train cost=1.462778704 test cost=1.479552042\n",
            "Epoch: 0069 train cost=1.462729837 test cost=1.479904852\n",
            "Epoch: 0070 train cost=1.462702536 test cost=1.479272977\n",
            "Epoch: 0071 train cost=1.462972325 test cost=1.479495918\n",
            "Epoch: 0072 train cost=1.463123710 test cost=1.479368640\n",
            "Epoch: 0073 train cost=1.462722113 test cost=1.478895953\n",
            "Epoch: 0074 train cost=1.462656378 test cost=1.479094922\n",
            "Epoch: 0075 train cost=1.462650900 test cost=1.478959805\n",
            "Epoch: 0076 train cost=1.462650175 test cost=1.478975938\n",
            "Epoch: 0077 train cost=1.462647153 test cost=1.478966271\n",
            "Epoch: 0078 train cost=1.462631225 test cost=1.478893672\n",
            "Epoch: 0079 train cost=1.462630910 test cost=1.478753238\n",
            "Epoch: 0080 train cost=1.463924875 test cost=1.480171909\n",
            "Epoch: 0081 train cost=1.462784419 test cost=1.479161767\n",
            "Epoch: 0082 train cost=1.462670432 test cost=1.479095173\n",
            "Epoch: 0083 train cost=1.462620465 test cost=1.479138353\n",
            "Epoch: 0084 train cost=1.462640048 test cost=1.479172059\n",
            "Epoch: 0085 train cost=1.462850694 test cost=1.479821035\n",
            "Epoch: 0086 train cost=1.462857049 test cost=1.479876047\n",
            "Epoch: 0087 train cost=1.462775665 test cost=1.479941851\n",
            "Epoch: 0088 train cost=1.462697300 test cost=1.479785857\n",
            "Epoch: 0089 train cost=1.462622495 test cost=1.479961377\n",
            "Epoch: 0090 train cost=1.462578588 test cost=1.479417449\n",
            "Epoch: 0091 train cost=1.462576804 test cost=1.479025338\n",
            "Epoch: 0092 train cost=1.463177819 test cost=1.479982987\n",
            "Epoch: 0093 train cost=1.462610721 test cost=1.479001541\n",
            "Epoch: 0094 train cost=1.462539916 test cost=1.479158794\n",
            "Epoch: 0095 train cost=1.462534701 test cost=1.479351459\n",
            "Epoch: 0096 train cost=1.462528601 test cost=1.479368678\n",
            "Epoch: 0097 train cost=1.462527778 test cost=1.479257174\n",
            "Epoch: 0098 train cost=1.462527277 test cost=1.479143121\n",
            "Epoch: 0099 train cost=1.462527068 test cost=1.479119196\n",
            "Epoch: 0100 train cost=1.462526898 test cost=1.479078905\n",
            "Accuracy: 0.9817\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAagAAAEWCAYAAAAzcgPFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3XmcXfP9x/HXO8lEVoksqsQSS5FY\nIh1bKRq6KKqLUhJL8PPTX1tUtaXVUrqgv2q1tKSkqlQQtH60jT2pWhOCEGqpZTRIokhIJJP5/P74\nnhsnk5k7d5Y792bm/Xw87mPmnvV7zr33fM75nu/5fhQRmJmZVZselS6AmZlZUxygzMysKjlAmZlZ\nVXKAMjOzquQAZWZmVckByszMqlKHByhJT0jaq6OX21EkbSRpsaSe1bReSWdKurLRsH9I2qETyna9\npH3bOO94Sbd2dJnau15Jd0s6Nvd+uKSnJPXthLJ5f1YRSR+QNFfSWh283Msl/bAjl1lOkj4q6ek2\nzluR42aHB6iIGB0Rd3f0cvPa88WIiJciYkBErOjocnXkeiUdACyKiEey90dJWpF9SQqvvXLTny3p\ncUn1ks5stKz9JN0j6U1Jr0q6VNLA3CTnAm3dn1dFxCfaMm97tGG9pwKXR8QSAEkHS7pX0ruS7s5P\nKOlDkv4sab6kNyRNk7Rlbrwk/VDSK5Leyg7eo3OL6Hb7sz0kvSBpn/Yup7nlRcRrwF3AcSXMOyD7\nbf21o8rTFtl37JuSnpG0RNJLkn7SmiArKSRtXngfEX+PiC2LzdOcSh03XcVXvY4H/tBo2H3Zl6Tw\nujs37lngW8AtTSxrEOmAuT6wNbAB8NPCyIh4EFhbUm0Hlr9qZD/qI4H8FeobwC+Ac5qYZTBwE7Al\n8AHgQeDPufFfBI4GPgoMAe4j91l10/1Z7a4C/ruE6b4AvAd8XNJ65S1SUb8kBdQjgIHAvsDewLUV\nLFOHktSrxYkiokNfwAvAPsCZpJ15BbAIeAKobTTdacCTwH+A3wF9snFHAfc0Wm4Am5M+tOXAMmAx\n8H/NlGMnYCbwNvAacH42fJNsWb2y9yOBGVkZbwcuAq5sNO1E4OWsnMcDOwKPAW8CF+bW2QM4HXgR\neD3b9kFF1js9W+9twIW59fYGlgAjcstebZ80s91XAme2MM3ngccbDfstcEaReY4Cns/K+y9gfFPl\nAj4BPA28Bfw628Zjc9P+A/h5tu+eBz6SDX8522dH5pY1KNuH87N9ejrQo5n1fhx4KlvvhY3Wuwfw\nbDPbdSxwdwv7a0j22Q3N3n8buDY3fjSwtDvvT9LJz02kwP8s8F+5cZcDP8y93wuoy/7/A9BA+r4v\nJp1kbZLt7+OAfwPzgFPaurxseC/gXWDjFj7rO4EfAQ/n15mN2yEbvgi4BphSKAewDnBztm//k/2f\n//3eTTpJvDcr1/8BQ0mB823gIWCTbNotgBXATo3WvyEpeI7L7YeLScePRdlntHE2bka2D9/J1ndI\nfj/ljsHfJB3L3gEuI52Q/ZX3j4frND5+Abtmyyy8lgIv5I6BpwLPAQtJMWBIo2UcA7wEzGjpeFbu\nK6jPkD7EwhnphY3Gjwc+CWwGfIj0gykqIiaRPtTzIl1FHNDMpBcAF0TE2tnymzvz+CPpDHkoKage\n3sQ0O5O+NIeQzrq/SwrCo4GDJe2ZTXdU9voYsCkwgNW3Ob/eWcAw4GzSGWnBFkBDRNQ1mmcHSQsk\n/VPS90o6A2naHqQThry5wPZNTSypP+mMbt+IGEg6CM5uYrphwFTSicdQ0oH1I40m25n0gxhK2gdT\nSAF/c2ACcKGkAdm0vyIdVDcF9iSdTU5sZr03kL4/w0g/jt1yk2yblaWt9gBejYiF2fspwGZZVWAN\n6bP7W6N5utv+nALUkQLVQcCPJY1ravvzIuJw0sHqgOz3fF5u9MdIv4VPAN8upRqwueVFRD0pcDb5\nmQBI2ph0EL8qex2RG9cb+BMpAA4BriNdbRX0IJ1kbwxsRAqQjX/7XyIdXzYgHZPuy+YZQvq+nJFN\ntzcpkDzYaNteBu4nnTwUjCcdP4aRvkNXZdPukY3fPtsP1zSz2V/Ilvch4ABScPoOMDzbphMazxAR\nK2tySIH5AeDqbPTXgM+Svl/rk4L1RY0WsSepJueTzZRppXIHqHsi4i+R6i3/wOpfjgsj4uWIeIN0\n1nJoB657ObC5pGERsTgi7m88gaSNSD/m70fEsoi4hxRIGzs7IpZGxK2kM42rI+L1iHgF+DvpzArS\nl+X8iHg+IhaTDixfahxIcuv9XkS8FxEzSGdUBYNJZzB5M4BtgHVJX6pDSWc/rSLp46QD6vcbjVqU\nrbc5DcA2kvpGxLyIaBzgAD4NPBERN2QHhF8Crzaa5l8R8bvsO3EN6azwrGw/3Eq6Mt48uxn7JeC0\niFgUES8AP6PpE4jCeqdGxHLSSUR+vU3tz5JIGkH6gZ2cGzwPuId0kF5CqvL7eqNZu83+lLQhKYB9\nO/udzAYuJXeAb6MfRMQ7EfE46UDe3uNDS5/J4cBjEfEkKeCOzjVS2gWoAX4REcsjYirpqgeAiFgY\nEddHxLsRsYh0PNuz0fJ/FxHPRcRbpEDwXETcnn221/H+cWQY6TvWlHnZ+IJbImJGRLxHOnHeNfs8\nSvWriHgtdyx7ICIeiYilwI25MjXnl6T9+t3s/fHAdyOiLivTmcBBjY6BZ2afa4v3L8sdoPJf6neB\nPo0K+nLu/xdJEbfVshZIhYYDhZubx5DOCp6S9JCk/ZuYdX3gjYh4t5kyFbyW+39JE+8LZ6jrZ9tR\n8CLpkvgDTaz3PxHxTqNpC/5DqndeKQt6/4qIhuwHexbpTLVkknYhnWUfFBH/bDR6IKmaCEkX5/bn\nd7JyHkL68s2TdIukrZpYxfrk9l+k6/rGV4GN913hJnZ+2ADSj7CG1ffnBiWuN/85rrY/SyFpOHAr\n8OuIuDo36vukE4wNgT7AD4A7JfXLTdOd9mfhd5Q/CWhu2a3RIceHnPxnkm9stFE2/gjevwJ5hVRl\nVqjZWB94JdsX+TKRLa+fpEskvSjpbdIJ5eBGrd5KPY4sAD7YzDZ8MBtfkP+cFpOqWFuzn0ot02ok\n/TfpivOwiGjIBm8M3Jg1yHqTdGW4glWPgU0dY5tU6UYS+Ui/Eam+GdJVysofexM3K1fpgj1SC6RC\nw4F9s2HPRMShpCuOc4GpWdVK3jxgSKMDS2vOPhr7N+kDKtgIqGfVD72w3nUalWej3P/PkhryFPuB\nB6BSC5adCd4EHB0RdzQxydbAowARcXxuf/44GzYtIj5O+oE8RbrH0tg8YERuncq/b6UFpKvgxvvz\nlWbWu/Jzy9ab/xwfI52slEzSOqTgdFNE/KjR6DHANdlZYn1EXE6q6hiVm6Y77c9/k35H+aCVX/Yq\nv2eg6O85p6TjQynLy06MN+f9zyTf2OglSR8hVSeeptTS9VVS9elh2bzzgA2yfZEvU8E3SI1qdo50\nW6FQxVbybzTnTmBDSTs12oYNSVdy+d9v/nMaQKou/DdlJumjpKrFAyPi7dyol0lV14Nzrz5ZwC8o\nOYVGpQPUVySNkDSEdIlYqCd9lHR5PUZSH9JlYt5rpHr0ZkmaIGl4FtnfzAY35KeJiBdJDSnOlNRb\n0q6keti2uhr4uqSR2Zflx6QDWX0z6/1Btt7d8+uNiGWkG5Qrqwgk7SvpA9n/WwHfI9eyTFJNtq96\nAL0k9SmcvUnahnSP5GsRka9KzNuTVO2wGqXnSA7MAup7pBujDU1MeguwraTPZj/qr7D6waMkWZXV\ntcCPJA3M7g+cTNMtx24hfV8+n633hEbrfZB0Nrsy4Evqme2vXkCPbH/VZOPWBqYB/4iIU5tY30PA\nF7P90kPS4aSrk2dz03Sb/Rnp3si9wE+y/bgdqQajsOzZwKclDclONk9qtL7mfs/fy65MRpPulRWO\nD21Z3k6kG/kv0rQjSY0NRpFOQMaQqtT7klrQ3Uc62Twh+619PltmwUDSFceb2fHsDNooq924GLhK\n0i7Zd3U0cD1we0Tcnpv805J2V7pHdjZwf/Z5QAnHybbIAuW1wBFN1MRcTPqObZxNO1zSgW1dV6UD\n1B9JZ6nPk27E/hBWfkBnkQ7Sz5Dq+/MuA0Zll5F/ambZnwKekLSY1GDiS83UeY4ntUpZmK3/GtJB\noy0mk+61zSC1zFpKumnYlMNIZ2hvkL7MVzQafwmr3h/YG3hM0jvAX0g3sX+cG/9b0g/kUFKwX5Kb\n/xukm56X5ao1Vt7zkLQjsDga3ZTN6UE6mP07K++ewJcbTxQRC0j3Y84j7c9RpEDc1v35NdLZ8vOk\n78AfSfu4ufWek613C1LrtsL4ZaQWTxNysx1O2ke/ITUXX8L7VzGfI1XhTWymKuhc0knUbNLJz9eB\nL0REofqoO+7PQ0mttP5NundxRu5A+gfS/nqB9HtvfMP+J8Dp2e/5lNzw6aSgfwfwv9k9tbYubzzp\n4Lma7ETlYNL9mFdzr39l6zoy2+bPkxpBvUGqor0ht5hfkILZAlJDhsaNZlrrq6T7eFeSTmD+RmoJ\n+IVG0/2RdPx4A/gwq34mZwK/z/bDwe0sT97epCq7qU0cTy4g1dTcKmkRaV/s3NYVadUq1c4j6QVS\ns9XbW5q2M0m6BngqItp8BtSBZfkH8NXIHtYt43quBy6LiL908HJ7kO6ZjI+Iuzpy2W0oy3CyBi2l\n3Jxt57q8P9u37E1IJ3g1jWsf2ri8dUnBbofs5n+XIOlyUmu/Fls/r6na2ky5y8jOdt8g/SA+ARxI\n0w9vdrqI2K3lqTpkPY3PytpM0idJzU6XkFoZinQWVVERMR9oqiFCOdbl/VlFIuJ10j1BW8OUrYpP\n0mRJr0ua08wkawMXS5otaY5SNz5Dsnm/rtSn3xxJV2eX4OWyHunSeTGpyeSXy33F0sXtSqquXUC6\nr/bZcl+xdHHen9Ztla2KT9IepIP+FRGxTQvTHgB8PSLGZTde7wFGRcQSSdcCf8laSpmZWTdRtiuo\nSA+fvlHi5Ify/pPIkKoe+2YtiPrRCc0mzcysulT8HpTSM0ifIrVaISJekfS/pO5KlgC35lrvNDX/\ncWS9FPfv3//DW221RlSLm5lVhVmzZi2IiOGVLkdTKh6gSPXq/4jU3VHhAckDSZ2pvglcJ2lCRDTZ\nc3KkvvkmAdTW1sbMmTM7p9RmZl2ApOaeDau4Sj8HBal/sHz13j6k/sXmR+oH7AZW7yDTzMy6uIoG\nKEmDSA8p5nPtvATskj1BLtJDYXMrUT4zM6ucslXxSbqa1JHgMEl1pKedawAiovBE9+dI95hWdpoa\nEQ9ImkrKu1IPPEJWhWdmZt1HxXqSKAffgzLrGpYvX05dXR1Ll3aZjh8qrk+fPowYMYKamppVhkua\nFRFVmf25GhpJmJmtoq6ujoEDB7LJJpsgtaVDcMuLCBYuXEhdXR0jR46sdHFKVg2NJMzMVrF06VKG\nDh3q4NRBJDF06NA17orUAcrMqpKDU8daE/enA5SZmVUlBygzs0YWLlzImDFjGDNmDOuttx4bbLDB\nyvfLli0raRkTJ07k6aefLnmdl156KSed1Dj3YvfmRhJmZo0MHTqU2bNnA3DmmWcyYMAATjnllFWm\niQgigh49mj7P/93vflf2cnZ1voIyMyvRs88+y6hRoxg/fjyjR49m3rx5HHfccdTW1jJ69GjOOuus\nldPuvvvuzJ49m/r6egYPHsypp57K9ttvz6677srrr79e8jqvvPJKtt12W7bZZhu+853vAFBfX8/h\nhx++cvgvf/lLAH7+858zatQotttuOyZMmFBssWsEX0GZWVU76STILmY6zJgx8ItftG3ep556iiuu\nuILa2vTo0DnnnMOQIUOor6/nYx/7GAcddBCjRo1aZZ633nqLPffck3POOYeTTz6ZyZMnc+qpp7a4\nrrq6Ok4//XRmzpzJoEGD2Geffbj55psZPnw4CxYs4PHHHwfgzTffBOC8887jxRdfpHfv3iuHrcl8\nBWVm1gqbbbbZyuAEcPXVVzN27FjGjh3L3LlzefLJJ1ebp2/fvuy7774AfPjDH+aFF14oaV0PPPAA\n48aNY9iwYdTU1HDYYYcxY8YMNt98c55++mlOOOEEpk2bxqBBgwAYPXo0EyZM4Kqrrlrtgdw1UTm7\nOpoM7A+83lTCQknfBMbnyrE1MDzXq3lPYCbwSkTsX65ymll1a+uVTrn0799/5f/PPPMMF1xwAQ8+\n+CCDBw9mwoQJTT5r1Lt375X/9+zZk/r6+naVYejQoTz22GP89a9/5aKLLuL6669n0qRJTJs2jenT\np3PTTTfx4x//mMcee4yePXu2a12VVM4rqMtJeZ6aFBE/jYgxETEGOA2YXghOmRNxJ7FmVsXefvtt\nBg4cyNprr828efOYNm1ahy5/55135q677mLhwoXU19czZcoU9txzT+bPn09E8MUvfpGzzjqLhx9+\nmBUrVlBXV8e4ceM477zzWLBgAe+++26Hlqezle0KKiJmSNqkxMlXyagraQSwH/Aj4OQOL5yZWQcY\nO3Yso0aNYquttmLjjTdmt912a9fyLrvsMqZOnbry/cyZMzn77LPZa6+9iAgOOOAA9ttvPx5++GGO\nOeYYIgJJnHvuudTX13PYYYexaNEiGhoaOOWUUxg4cGB7N7GiytpZbBagbm6qii83TT+gDtg8V703\nFfgJMBA4pVgVXz6j7kYbbfThF1+s2txbZlaiuXPnsvXWW1e6GF1OU/u1mjuLrYZGEo0z6hbuW80q\nZeaImBQRtRFRO3x4VWYtNjOzNqiGANU4o+5uwGckvQBMAcZJajLdu5mZdV1Vl1E3Ik6LiBERsQkp\neN0ZEWv+E2dmZtYqVZdR18zMDMrbiu/QEqa5nNQcvbnxdwN3d1SZzMxszVEN96DMzMxW4wBlZtZI\nR6TbAJg8eTKvvvpqk+MmTJjAn/70p44qcpfkzmLNzBopJd1GKSZPnszYsWNZb731OrqI3YKvoMzM\nWuH3v/89O+20E2PGjOF//ud/aGhoaDL9xTXXXMPs2bM55JBDSr7yamho4OSTT2abbbZh2223Xdmr\nxCuvvMLuu+/OmDFj2Gabbbj33nubTbnRlfgKysyqWxXl25gzZw433ngj9957L7169eK4445jypQp\nbLbZZqulvxg8eDC/+tWvuPDCCxkzZkxJy7/uuuuYO3cujz76KPPnz2fHHXdkjz324Morr+SAAw7g\n29/+NitWrGDJkiXMmjWryZQbXYmvoMzMSnT77bfz0EMPUVtby5gxY5g+fTrPPfdcs+kvWuuee+7h\n0EMPpWfPnqy33nrsvvvuzJw5kx133JFLL72UH/zgB8yZM4cBAwZ02Dqrma+gzKy6VVG+jYjg6KOP\n5uyzz15tXFPpLzrKuHHjuPvuu7nllls44ogj+Na3vsX48ePLus5q4CsoM7MS7bPPPlx77bUsWLAA\nSK39XnrppSbTXwAMHDiQRYsWlbz8j370o0yZMoWGhgZee+01/vGPf1BbW8uLL77Ieuutx3HHHcfE\niRN55JFHml1nV+IrKDOzEm277bacccYZ7LPPPjQ0NFBTU8PFF19Mz549V0t/ATBx4kSOPfZY+vbt\ny4MPPrhK4kKAY489lq9+9asAjBw5kunTp3P//fez3XbbIYnzzz+fddddl8mTJ3P++edTU1PDwIED\n+cMf/sDLL7/c5Dq7krKl22hPRl1JnwIuAHoCl0bEOaWss7a2NmbOnNkh5TezynG6jfJwuo33XU4b\nMupmqd4vAvYFRgGHShpVxnKamVkVKluAiogZwBstTpjkM+ruBDwbEc9HxDJSyo0Dy1BEMzOrYhVv\nJJFl1P0UcH02aAPg5dwkddkwM+tGypntuztaE/dnxQMUjTLqtpak4yTNlDRz/vz5HVw0M6uEPn36\nsHDhwjXyoFqNIoKFCxfSp0+fShelVaqhFV/jjLqvABvm3o/IhjUpIiYBkyA1kihHAc2sc40YMYK6\nujp80tlx+vTpw4gRIypdjFapaIDKZdTNZ8x9CNhC0khSYPoScFgFimdmFVJTU8PIkSMrXQyrsKrL\nqBsR9ZK+CkwjNTOfHBFPlKucZmZWncr2HFQl+DkoM7PWWaOfg5J0nqS1JdVIukPSfEkTWprPzMys\nPUppxfeJiHib1CvEC8DmwDfLWajOJqWXmZlVj1ICVOE+1X7AdRHxVhnLY2ZmBpTWSOJmSU8BS4Av\nSxoOLC1vsczMrLtr8QoqIk4FPgLURsRy4B26WNdDv+NITuT8ShfDzMxySmkk8UVgeUSskHQ6cCWw\nftlL1om+wA3sxEOVLoaZmeWUcg/qexGxSNLuwD7AZcBvyluszrWEvvThvUoXw8zMckoJUCuyv/sB\nkyLiFqB3kenXOEvpQx/fVjMzqyqlBKhXJF0CHAL8RdJaJc63xnCAMjOrPqUEmoNJ3Q59MiLeBIbQ\nxZ6Deo/e9GVJpYthZmY5pbTiexd4Dvhk1kfeuhFxa0vzSZos6XVJc4pMs5ek2ZKekDQ9N3ywpKmS\nnpI0V9KuJW5PmyxjLfrxbjlXYWZmrVRKK74TgauAdbPXlZK+VsKyL6dIyndJg4FfA5+JiNHAF3Oj\nLwD+FhFbAdsDc0tYX5sto8YBysysypTyoO4xwM6FHsclnQvcB/yq2EwRMUPSJkUmOQy4ISJeyqZ/\nPVv+IGAP4Khs+DJgWQnlbLPl9GYAi+nVC+rry7kmMzMrVSn3oMT7LfnI/u+Inus+BKwj6W5JsyQd\nkQ0fCcwHfifpEUmXSurfbOE6IKPucnoxgMWsWNHytGZm1jlKCVC/Ax6QdKakM4H7Sc9CtVcv4MOk\n5uufBL4n6UPZ8LHAbyJiB1LPFac2t5CImBQRtRFRO3z48DYVpJ5e9Ocd+tCmrPNmZlYGpTSSOB+Y\nCLyRvSZGxC86YN11wLSIeCciFgAzSPeb6oC6iHggm24qKWCVzQp60oNgDI+XczVmZtYKzd6DkjQk\n9/aF7LVyXES093Ljz8CFknqRHvzdGfh5RLwq6WVJW0bE08DewJPtXFdRK+gJwAd5tZyrMTOzVijW\nSGIWELx/v6mQelfZ/5sWW3BLKd8jYq6kvwGPAQ3ApRFRaJL+NeAqSb2B50lXcGXTkG3iMN4s52rM\nzKwVmg1QETGyPQuOiENLmOanwE+bGD4b6MQUxKmm003NzcyqR5fqsqitRAOAe5MwM6siDlBADcsB\n6OMAZWZWNRyggL4sBmCt8j4PbGZmrVBKV0c/kzS6MwpTKZvxTwD3aG5mVkVKuYKaC0yS9ICk47Ou\niLqUDUk9UPT2FZSZWdUo5UHdSyNiN+AIYBPgMUl/lPSxcheu0/RMz0G5is/MrHqUdA9KUk9gq+y1\nAHgUOFnSlDKWrfPU17OMGtZyFZ+ZWdVosTdzST8HDgDuAH4cEQ9mo86V9HQ5C9eZ3qUfa/FepYth\nZmaZUtJtPAacXki30chOHVyeillCX1fxmZlVkVKq+C4nZdM9P2vR97nCiIh4q7mZ2ppRV9KGku6S\n9GQ2/MRWbE+bLaWPH9Q1M6sipQSoi4DjgceBOcB/S7qohPkup20ZdeuBb0TEKGAX4CuSRpWwvnZZ\nSh/fgzIzqyKlVPGNA7aOiACQ9HvgiZZmamtG3YiYB8zL/l8kaS6wAWXu0fw91qKfr6DMzKpGKVdQ\nzwIb5d5vmA1rr+Yy6q6UBbgdgAcaj8tN0+6MupAClKv4zMyqRylXUAOBuZIKrfd2BGZKugkgIj7T\njnV/mJTvqS9wn6T7I+KfAJIGANcDJ0XE280tJCImAZMAamtro7npWrKM3u7N3MysipQSoL5fpnXX\nAQuz1oHvSCpk1P2npBpScLoqIm4o0/pXsZwaBrCYXXaB++/vjDWamVkxLQaoiJgu6QOkKyeABwv3\ni9qpyYy6kgRcBszN0s13ikKAeqDZykQzM+tMpXQWezDwIKmV3cHAA5IOKmG+q4H7gC0l1Uk6JuvL\n73iAiJgLFDLqPsj7GXV3Aw4HxmVN0GdL+nQbt69ky+lJX5ayIU+Ve1VmZlaCUqr4vgvsWLhqkjQc\nuB2YWmymtmbUjYh7eD/NfKepT9no2Y65pB6dzMyskkppxdejUZXewhLnW6OsIHUYO5wFFS6JmZlB\naVdQf5M0Dbg6e38I8JfyFakyGrKYO4hFFS6JmZlBaY0kvinp88Du2aBJEXFjeYvV+VZktYp93ZuE\nmVlVKBqgsjQbt0fEx4BOae5dKSI9QtXHD+uamVWFoveSImIF0NAVs+g21isLUO7R3MysOpRyD2ox\n8Lik24CVKTci4oSylaoCarJeJJz23cysOpQSoG5g9eq9NncpVK024CUAJy00M6sSpQSowRFxQX5A\nZ+Vo6kwjs87SeztAmZlVhVKeZzqyiWFHdXA5Km7IzjsDsBbLK1wSMzODIgFK0qGS/g8YKemm3Osu\n4I2WFtxSRt0sm+5bue6Mvp8b9/Usm+4cSVdL6tOWjWuV++9niZMWmplVjWJVfPeSEgcOA36WG76I\n1H9eSy4HLgSuKDLN3yNi//wASRsAJwCjImKJpGuBL2XLK6t36O9WfGZmVaLZABURLwIvAru2ZcEl\nZNQtphfQV9JyoB/w7zYup1WW0Jc+voIyM6sKpfRm/nlJz2TVcW9LWiSp2QSCrbSrpEcl/VXSaICI\neAX4X+Al0hXcWxFxa5HydUhGXYCl9HErPjOzKlFKI4nzgM9ExKCIWDsiBkbE2h2w7oeBjSNie+BX\nwJ8AJK0DHAiMBNYH+kua0NxCImJSRNRGRO3w4cPbVaAl9PEVlJlZlSglQL2W5W7qUBHxdkQszv7/\nC1AjaRiwD/CviJgfEctJz2B9pKPX35RlrEVfd3VkZlYVSnkOaqaka0hXOCvrv9qbil3SeqTgF5J2\nIgXLhaSqvV0k9QOWAHsDM9uzrlK9x1oMdG/mZmZVoZQAtTbwLvCJ3LCghc5js4y6ewHDJNUBZ0DK\nChgRFwMHAV+WVE8KRF+KiCBl7J1KqgKsBx4BJrVim9rsPXrT//3enMzMrIKUYkLXUFtbGzNntv1i\na5o+wU48xPW//Q/HHtuBBTMzq1KSZkVEbaXL0ZRiD+pem/v/3Ebjmm1VtyZbRm8GsJj/+q9OadVu\nZmZFFGsksUXu/483Gte+5nISqhgqAAAWk0lEQVRVqp5e1FDPx2my8wszM+tExQJUsbq/rlMvmLM8\n3SJjBP+qcEnMzKxYI4l+knYgBbG+2f/KXn07o3CdbUW2OwbTUc8hm5lZWxULUPOA87P/X839X3jf\n5dRnF5T9/LCumVnFFeuL72OdWZBq0JDtDueEMjOrvFJ6kug2VmR/+zhAmZlVnANUTs/sb42TFpqZ\nVZwD1CpSLqjezgllZlZxpaTb2E1S/+z/CZLOl7RxCfO1OaNuNr6npEck3VzqxrTX2lmiYCctNDOr\nvFKuoH4DvCtpe+AbwHMUz5JbcDnwqRam+XtEjMleZzUadyLQ4b2oF7Nxlii4xgHKzKziSglQ9Vkn\nrgcCF0bERcDAlmaKiBmQXZK0kqQRwH7ApW2Zv63G/vZsGpCvoMzMqkApAWqRpNOACcAtknqQ9Ure\nAVbLqJv5BfAtoKGlBXRkRl2OPZZ36O9GEmZmVaCUAHUIKQ/UMRHxKjAC+GkHrLu5jLr7A69HxKxS\nFtKRGXUhZdWF4LLL2r0oMzNrh5KuoIALIuLvkj4EjAGubu+Ki2TU3Q34jKQXgCnAOElXtnd9pZrH\nB9mL6Zx57HOdtUozM2tCKQFqBrCWpA2AW4HDSQ0g2kXSepKU/b8yo25EnBYRIyJiE+BLwJ0RMaG9\n6ytVAz0ZyhscyvWdtUozM2tCKRl1FRHvSjoG+HVEnCfp0RZnantG3Yragdm8zUA+wr2VLoqZWbdW\nUoCStCswHjgmG9bilVdEHNrC+AuBC1uY5m7g7hLK2HEieEy7M4472Uu3cXc0ToVlZmadoZQqvpOA\n04AbI+IJSZsCd5W3WJX1OsNZm0V8mi6ZONjMbI3Q4hVUREwHpksaIGlARDwPnFD+olXOUB5lAUPZ\ngUcqXRQzs26rlK6OtpX0CPAE8KSkWY2eWepy9ozneYgd2Y17OVPfqXRxzMy6pVKq+C4BTo6IjSNi\nI1J3R78tb7Eq70m2pB9L2Jh/V7ooZmbdUikBqn9ErLznlDVc6F+2ElWJWxhMHetTyyyYPLnSxTEz\n63ZKCVDPS/qepE2y1+nA8+UuWKXdGWdyLQezLXO445g/VLo4ZmbdTikB6mhgOHADcD0wLBvW5f2Q\nr/IQtYxlNm+nZ4rNzKyTFG3FJ6kn8N2I6NKt9przRmzGeJ3I7zmKuWzDtpUukJlZN1L0CioiVgC7\nd1JZqtL9m07gUo5lW+bwqtrfGa2ZmZWmlCq+RyTdJOlwSZ8vvFqaqT0ZdSV9StLTkp6VdGortqfD\nPfcc/IwTeYkN6c0KcFWfmVmnKKWroz7AQmBcbliQ7kkVczmpK6Ni2Xf/HhH75wdk1YoXAR8H6oCH\nJN0UEU+WUNayeCa25pO6hOv5IovpzwAJKt9toJlZl1ZKTxIT27LgiJghaZM2zLoT8GzWYwWSppCy\n+VYsQAHcyr4cwhRu4AssZS36OEiZmZVVs1V8kn4q6b+bGP7fks7poPU3lVF3A+Dl3DR12bDmytlx\nGXWLiIC/sD+HcRW9qGcFPVzdZ2ZWRsXuQY0DJjUx/LfA/k0Mb60mM+q2Vkdn1C2+LriBgziMP7Kc\nGgIcpMzMyqRYgFqrqfxMEdEAtPuoXCSj7ivAhrlJR2TDqkIEXMfBbMMcnmKrNFByoDIz62DFAtQS\nSVs0HpgNW9LeFTeXURd4CNhC0khJvUlZdW9q7/o6UkMDPMfmbMdj/IqvsDKKO1CZmXWYYgHq+8Bf\nJR2V9Wi+raSJwC3ZuKKyjLr3AVtKqpN0jKTjJR2fTXIQMCfLzvtLsoy6EVEPfBWYBswFro2IJ9q+\niR2v0D6inhpO4EI25xleY/iqEzhYmZm1i4plWZe0DfBNYJts0BzgfyPi8U4oW6vV1tbGzJkzO3Wd\n+Rj0Cf7K3/h08/WfbvVnZlVG0qyIqK10OZpStJl5RMwBjuyksqyRIt4PUreyLz0IaniHZQxYfeLm\nrqgcuMzMVlNKTxLWggi4/PL33y+nPyIQAVcUe045k68SbMvruuvKtm1mZpVStIpvTVOJKr7GmrtI\nOvpouOyyFiayNdcHPwh//jPsuOP7w44+Gq66CjbeGK6+Gm66Cc46q3Jl7NcvlXPs2FSOrbZadfwz\nz8CHPlT+ctTUwLhxcMYZcP758Mgj8NZb8IEPwGabwXbbwbvvvj99r6yiZ+BA2HTTNO96670//vXX\nYaON4L33Vl1Pr14wZAj07p22vaEhvZYvh/r6NE1DA7z55urztlaPHtCnDwwdCltuCZtsAjffnMrW\n0AA9e6Z9P2FC2oaXX4ZXXknbvWgRvP12WkZNTVpev34wbBj07w+DB8P666dh/fvDOuukaddaC4YP\nT//37g0jR7ap6NVcxdemACWpd0QsK0N52qUaAlRBsRjU4i7fe2+4884OLY+ZdWE9esCKFW2atZoD\nVItdHUm6GzgqIl7I3u9Eelh3+7KWbA1XCEL9+696MgirBq8rr4Tx4xvNfMcdZS1bxU2bls5i92/h\nee9ly9KZYVMmTly1XhVg0KB0Jrom1AoMGpSuHiJgyhS4/fZKl6h9Bg5MVwKVduml6XX//a2fd+21\n0xXcoEHph9uvX7pyGTw4/Whfeim9Fi1K040ala7Ebrop9Sq9eHG6Wip8/3r1SldSRx8N118PTzwB\nS5euWj3fq1e6aqqpSd/1mpo0vGfP9Fq+/P0rpPr6NGzw4LSOmhoYMSKts3/XTHLe4hWUpE8CF5Ca\ngm8A7AscGxEPl794rVNNV1BNKaVm76qr4LDDyl8WMzOo7iuoFhtJRMQ04HhSkDoa+HQ1Bqc1QUR6\nNTQ0P8348U23gzj++ObnMTPriloMUJK+R+orbw/gTOBuSfuVuVxdWuFB38Lrv/6r5XkuuaT5Rnx9\n+5a/zGZmna2UZuZDgZ0i4r6IuAT4JHBSeYvVvUyatGrAikhVfaVqXK1d6uvGG8u3TWZm7VVKFd9J\nEbEk9/7FiPh4S/O1lFE3N92OkuolHZQbtpGkWyXNlfRkG/NKrdEOO2z1oFV4ffnLHbOOz3++/Y9g\nVfNr661X3+amqljPPXf1eceMKb7vvvWt6rtX+MQTcNJJ8GQHZE5raGhzo7CKOfvs1J5g000rXRLr\nKKU0khgOfBsYRcquC0BEjGt2pjTfHsBi4IqI2KaZaXoCtwFLgckRMTUbfjfwo4i4TdIAoCEi3m1q\nGXnV3kiiM331q3DRRZUuhZm1RY9curlCY7/C8EKVfq9esO666f8BA+Dee9u2rjW6kQRwFanT1pHA\nD4AXSD2OFxURM4A3Wpjsa8D1wOuFAZJGAb0i4rZsOYtLCU62qgsvbP4KrCNejZ/x3GqrVcebWdsV\nrmBXrEity5cuTa9334WFC9Prtddgzpz0mjWr0iUujxafgwKGRsRlkk6MiOnAdEktBqiWSNoA+Bzw\nMSD3+D0fAt6UdAMpKN4OnBoRTVY4SDoOOA5go402am+xrERz5xYf7yBlZu1VyhXU8uzvPEn7SdoB\nGNIB6/4F8O0sAWJeL+CjwCmkwLUpcFRzC+nMjLpmZtZ5SrmC+qGkQcA3SM3N1wa+3gHrrgWmZDkL\nhwGfllQP1AGzI+J5AEl/AnYBLmtuQWZm1vW0GKAi4ubs37dI1XEdIiJW9mwo6XLg5oj4U9ZwYrCk\n4RExHxgHuOWDmVk302yAkvTLYjNGxAnFxmcZdfcChkmqA84AarJ5Ly6y3BWSTgHuyFLCzyL1/Wdm\nZt1IsSuo40kZdK8F/g3NJ4ptSkQc2oppj2r0/jZgu9asz8zMupZiAeqDwBeBQ4B64BpgakS82RkF\nMzOz7q3ZVnwRsTAiLo6IjwETgcHAk5IO77TSmZlZt1VKPqixwKHAx4G/ku4JmZmZlVWxRhJnAfuR\nepGYApwWEfWdVTAzM+veil1BnQ78i5Q5d3vgx9kzSwIiItyIwczMyqZYgBpZZJyZmVlZNRugIuLF\nziyImZlZXil98ZmZmXU6BygzM6tKLQYoSSeWMqyJadqUUVfSGEn3SXpC0mOSDmlpXWZm1vWUcgV1\nZBPDjiphvsuBTxWbIOsY9lzg1tzgd4EjImJ0Nv8vJA0uYX1mZtaFFHsO6lDgMGCkpJtyowbScqZc\nImKGpE1amKyQUXdlwsKI+Gfu/39Leh0YDriLJTOzbqRYM/OHgXmkXE0/yw1fBDzW3hUXyaibn2Yn\noDfwXJHlOKOumVkXVCxAXR0RYyU9l6V672grM+pmDwCvQtIHgT8ARzaRdXeliJgETAKora11onEz\nsy6iWIDqLekwYFdJn288MiJuaOe6m8yomyUtXBu4BfhuRNzfzvWYmdkaqKV8UONJvZgf0GhcAO0K\nUEUy6vYGbgSuiIip7VmHmZmtuYr1JHEPcI+kmRFxWWsX3NaMusDBwB7AUElHZcOOiojZrS2DmZmt\nuRRR/LZNdkVzPCloAEwHLo6I5WUuW6vV1tbGzJkzK10MM7M1hqRZEVFb6XI0pcV8UMCvSVc+v87e\nHw78Bji2XIUyMzMrJUDtGBHb597fKenRchXIzMwMSutJYoWkzQpvJG0KrChfkczMzEq7gvomcJek\n50nJCjcGJpa1VGZm1u21GKAi4g5JWwBbZoOejoj3ylssMzPr7pqt4st6GV8PIAtIY4CzgZ9KGtJJ\n5TMzs26q2D2oS4BlAJL2AM4BrgDeIutayMzMrFyKVfH1jIhCr+WHAJMi4nrgekl+aNbMzMqq2BVU\nT0mFALY3cGduXCmNK8zMzNqsWIC6Gpgu6c/AEuDvAJI2J1XztailrLqSDsyy5s6WNFPS7rlx52VZ\ndedK+qWa6vLczMy6rGYDVET8CPgGKTPu7vF+n0g9SIkGS3E5xbPq3gFsHxFjgKOBSwEkfQTYDdgO\n2IaUL2rPEtdpZmZdQNGquqZSXeQz3rakpay6EbE497Y/qZd0sr99SMkKRepq6bVS12tmZmu+UnqS\nKCtJn5P0FCn/09EAEXEfcBcpo+88YFpEzK1cKc3MrLNVPEBFxI0RsRXwWdJzVoX7XFsDI4ANgHGS\nPtrU/JKOy+5fzZw/f35nFdvMzMqs4gGqICJmAJtKGgZ8Drg/IhZn1YB/BXZtZr5JEVEbEbXDhw/v\nxBKbmVk5VTRASdq80DpP0lhgLWAh8BKwp6RekmpIDSRcxWdm1o2U9XmmErLqfgE4QtJyUlP2QyIi\nJE0FxgGPkxpM/C0i/q+cZTUzs+rSYkbdNYkz6pqZtU41Z9StmntQZmZmeQ5QZmZWlRygzMysKjlA\nmZlZVXKAMjOzquQAZWZmVckByszMqpIDlJmZVSUHKDMzq0oOUGZmVpXKGqDak/I9G7+2pDpJF5az\nnGZmVn3KfQV1OW1I+Z5zNjCjPEUzM7NqVtYAleV4eqPI+MXxfm+1+ZTvSPow8AHg1nKW0czMqlPF\n70E1lfJdUg/gZ8ApJczvjLpmZl1QxQNUUynfgf8B/hIRdSXM74y6ZmZdUFkTFrZGRMyQVEj5vivw\nUUn/AwwAektaHBGnVraUZmbWWSoaoCRtDjyXZdFdmfI9IsbnpjkKqHVwMjPrXqoy5Xs5y2RmZmsG\np3w3M+vGnPLdzMyslRygzMysKjlAmZlZVXKAMjOzquQAZWZmVckByszMqpIDlJmZVSUHKDMzq0oO\nUGZmVpWqNqOupCMlPZO9jixnOc3MrPpUZUZdSUNI/fbtDOwEnCFpnfIW1czMqkm1ZtT9JHBbRLwR\nEf8BbqN4oDMzsy6m4vmgJH0O+AmwLrBfNngD4OXcZHXZsKbmPw44Lnu7WNLTbSzKMGBBG+ddU3XH\nbYbuud3dcZuhe253a7d543IVpL0qHqAi4kbgRkl7kDLq7tPK+ScBk9pbDkkzq7VH33LpjtsM3XO7\nu+M2Q/fc7q60zVXTii+rDixk1H0F2DA3ekQ2zMzMuomKBihJm0tS9v/KjLrANOATktbJGkd8Ihtm\nZmbdRLVm1H1D0tnAQ9mizoqIZhtbdJB2VxOugbrjNkP33O7uuM3QPbe7y2xzl8qoa2ZmXUfV3IMy\nMzPLc4AyM7Oq1O0DlKRPSXpa0rOSTq10ecpF0oaS7pL0pKQnJJ2YDR8i6basS6nbumKPHZJ6SnpE\n0s3Z+5GSHsg+82sk9a50GTuapMGSpkp6StJcSbt29c9a0tez7/YcSVdL6tMVP+umupBr7rNV8sts\n+x/LGqOtMbp1gJLUE7gI2BcYBRwqaVRlS1U29cA3ImIUsAvwlWxbTwXuiIgtSF1PdcUgfSIwN/f+\nXODnEbE58B/gmIqUqrwuAP4WEVsB25O2v8t+1pI2AE4AaiNiG6An8CW65md9Oav3rNPcZ7svsEX2\nOg74TSeVsUN06wBF6ufv2Yh4PiKWAVOAAytcprKIiHkR8XD2/yLSAWsD0vb+Ppvs98BnK1PC8pA0\ngtRDSaGfRwHjgKnZJF1xmwcBewCXAUTEsoh4ky7+WZNaJfeV1AvoB8yjC37WzXQh19xneyBwRST3\nA4MlfbBzStp+3T1AldylUlciaRNgB+AB4AMRMS8b9SrwgQoVq1x+AXwLaMjeDwXejIj67H1X/MxH\nAvOB32VVm5dK6k8X/qwj4hXgf4GXSIHpLWAWXf+zLmjus12jj3HdPUB1O5IGANcDJ0XE2/lx2TNo\nXea5A0n7A69HxKxKl6WT9QLGAr+JiB2Ad2hUndcFP+t1SFcLI4H1SZ1Pd8sOprvSZ9vdA1S36lJJ\nUg0pOF0VETdkg18rXPJnf1+vVPnKYDfgM5JeIFXfjiPdmxmcVQNB1/zM64C6iHggez+VFLC68me9\nD/CviJgfEcuBG0iff1f/rAua+2zX6GNcdw9QDwFbZC19epNuqt5U4TKVRXbv5TJgbkScnxt1E1BI\nCHkk8OfOLlu5RMRpETEiIjYhfbZ3RsR44C7goGyyLrXNABHxKvCypC2zQXsDT9KFP2tS1d4ukvpl\n3/XCNnfpzzqnuc/2JlJvPZK0C/BWriqw6nX7niQkfZp0n6InMDkiflThIpWFUrbivwOP8/79mO+Q\n7kNdC2wEvAgc3AndSnU6SXsBp0TE/pI2JV1RDQEeASZExHuVLF9HkzSG1DCkN/A8MJF0QtplP2tJ\nPwAOIbVYfQQ4lnS/pUt91vku5IDXSF3I/YkmPtssWF9Iqu58F5gYETMrUe626PYByszMqlN3r+Iz\nM7Mq5QBlZmZVyQHKzMyqkgOUmZlVJQcoMzOrSg5QZlVA0l6F3tbNLHGAMjOzquQAZdYKkiZIelDS\nbEmXZLmmFkv6eZaL6A5Jw7Npx0i6P8vDc2MuR8/mkm6X9KikhyVtli1+QC6H01XZQ5Zm3ZYDlFmJ\nJG1N6qlgt4gYA6wAxpM6Jp0ZEaOB6aQn+wGuAL4dEduRevAoDL8KuCgitgc+Qup9G1IP8yeRcpNt\nSupLzqzb6tXyJGaW2Rv4MPBQdnHTl9QpZwNwTTbNlcANWU6mwRExPRv+e+A6SQOBDSLiRoCIWAqQ\nLe/BiKjL3s8GNgHuKf9mmVUnByiz0gn4fUSctspA6XuNpmtr/2H5PuJW4N+ndXOu4jMr3R3AQZLW\nBZA0RNLGpN9Rocfsw4B7IuIt4D+SPpoNPxyYnmUzrpP02WwZa0nq16lbYbaG8BmaWYki4klJpwO3\nSuoBLAe+QkoIuFM27nXSfSpIaQ8uzgJQoUdxSMHqEklnZcv4Yiduhtkaw72Zm7WTpMURMaDS5TDr\nalzFZ2ZmVclXUGZmVpV8BWVmZlXJAcrMzKqSA5SZmVUlBygzM6tKDlBmZlaV/h/51PQ79OfxfwAA\nAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_kK4oMJm8YgD",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "e962a27c-605a-47f8-9a7c-0ec87eda29e4"
      },
      "source": [
        "# Plot weight distriubtion at the end of training.\n",
        "\n",
        "import seaborn as sns\n",
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEMCAYAAADAqxFbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nnd9qWpzdbQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GhmMG9KkDE3m",
        "colab_type": "text"
      },
      "source": [
        "**Model 1 with Gradient Decent Optimizer**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p3Lr3d2QCx6u",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2150
        },
        "outputId": "b7897225-4d5e-4a0f-d99e-c9fd93dfe3f3"
      },
      "source": [
        "# We can now launch the model in an InteractiveSession\n",
        "\n",
        "# Note: make sure you initialize variables\n",
        "\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use GradientDescentOptimizer\n",
        "            _, c, w = sess.run([optimizer_sgdc, cost_sgd, weights_sgd], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_sgd, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(2, 2.6, step=0.05), \"input-sigmoid(512)-sigmoid(128)-sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(2, 2.6, step=0.05), \"input-sigmoid(512)-sigmoid(128)-sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(y_sgd,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 0001 train cost=2.321564540 test cost=2.321150202\n",
            "Epoch: 0002 train cost=2.318521373 test cost=2.318157343\n",
            "Epoch: 0003 train cost=2.315810137 test cost=2.315517844\n",
            "Epoch: 0004 train cost=2.313424408 test cost=2.313203239\n",
            "Epoch: 0005 train cost=2.311333858 test cost=2.311149142\n",
            "Epoch: 0006 train cost=2.309496613 test cost=2.309346313\n",
            "Epoch: 0007 train cost=2.307867411 test cost=2.307755907\n",
            "Epoch: 0008 train cost=2.306409356 test cost=2.306304021\n",
            "Epoch: 0009 train cost=2.305092844 test cost=2.305030156\n",
            "Epoch: 0010 train cost=2.303899262 test cost=2.303813533\n",
            "Epoch: 0011 train cost=2.302816279 test cost=2.302718572\n",
            "Epoch: 0012 train cost=2.301831965 test cost=2.301750970\n",
            "Epoch: 0013 train cost=2.300937148 test cost=2.300834191\n",
            "Epoch: 0014 train cost=2.300127413 test cost=2.300014296\n",
            "Epoch: 0015 train cost=2.299395790 test cost=2.299252249\n",
            "Epoch: 0016 train cost=2.298738532 test cost=2.298571920\n",
            "Epoch: 0017 train cost=2.298147681 test cost=2.297952731\n",
            "Epoch: 0018 train cost=2.297613135 test cost=2.297386894\n",
            "Epoch: 0019 train cost=2.297126886 test cost=2.296877471\n",
            "Epoch: 0020 train cost=2.296680126 test cost=2.296406294\n",
            "Epoch: 0021 train cost=2.296265528 test cost=2.295964683\n",
            "Epoch: 0022 train cost=2.295873640 test cost=2.295556330\n",
            "Epoch: 0023 train cost=2.295501189 test cost=2.295161662\n",
            "Epoch: 0024 train cost=2.295142510 test cost=2.294777470\n",
            "Epoch: 0025 train cost=2.294794208 test cost=2.294423347\n",
            "Epoch: 0026 train cost=2.294453392 test cost=2.294064053\n",
            "Epoch: 0027 train cost=2.294117421 test cost=2.293703821\n",
            "Epoch: 0028 train cost=2.293784975 test cost=2.293361636\n",
            "Epoch: 0029 train cost=2.293453683 test cost=2.293002984\n",
            "Epoch: 0030 train cost=2.293125329 test cost=2.292677001\n",
            "Epoch: 0031 train cost=2.292795767 test cost=2.292337239\n",
            "Epoch: 0032 train cost=2.292465594 test cost=2.291993745\n",
            "Epoch: 0033 train cost=2.292134398 test cost=2.291653659\n",
            "Epoch: 0034 train cost=2.291803384 test cost=2.291301772\n",
            "Epoch: 0035 train cost=2.291469642 test cost=2.290964771\n",
            "Epoch: 0036 train cost=2.291134550 test cost=2.290616875\n",
            "Epoch: 0037 train cost=2.290797670 test cost=2.290272993\n",
            "Epoch: 0038 train cost=2.290458977 test cost=2.289919460\n",
            "Epoch: 0039 train cost=2.290117664 test cost=2.289564238\n",
            "Epoch: 0040 train cost=2.289774823 test cost=2.289207489\n",
            "Epoch: 0041 train cost=2.289429480 test cost=2.288859224\n",
            "Epoch: 0042 train cost=2.289081226 test cost=2.288499696\n",
            "Epoch: 0043 train cost=2.288731119 test cost=2.288138711\n",
            "Epoch: 0044 train cost=2.288378189 test cost=2.287775103\n",
            "Epoch: 0045 train cost=2.288021871 test cost=2.287411599\n",
            "Epoch: 0046 train cost=2.287663919 test cost=2.287043044\n",
            "Epoch: 0047 train cost=2.287302453 test cost=2.286673053\n",
            "Epoch: 0048 train cost=2.286937976 test cost=2.286292459\n",
            "Epoch: 0049 train cost=2.286570187 test cost=2.285922900\n",
            "Epoch: 0050 train cost=2.286199627 test cost=2.285542485\n",
            "Epoch: 0051 train cost=2.285826515 test cost=2.285151377\n",
            "Epoch: 0052 train cost=2.285449222 test cost=2.284770027\n",
            "Epoch: 0053 train cost=2.285069395 test cost=2.284380173\n",
            "Epoch: 0054 train cost=2.284685209 test cost=2.283981590\n",
            "Epoch: 0055 train cost=2.284298151 test cost=2.283589152\n",
            "Epoch: 0056 train cost=2.283907386 test cost=2.283180366\n",
            "Epoch: 0057 train cost=2.283512267 test cost=2.282781885\n",
            "Epoch: 0058 train cost=2.283113493 test cost=2.282372105\n",
            "Epoch: 0059 train cost=2.282710511 test cost=2.281963414\n",
            "Epoch: 0060 train cost=2.282304469 test cost=2.281535757\n",
            "Epoch: 0061 train cost=2.281893313 test cost=2.281127125\n",
            "Epoch: 0062 train cost=2.281479163 test cost=2.280689754\n",
            "Epoch: 0063 train cost=2.281060153 test cost=2.280272335\n",
            "Epoch: 0064 train cost=2.280636574 test cost=2.279828439\n",
            "Epoch: 0065 train cost=2.280208695 test cost=2.279384778\n",
            "Epoch: 0066 train cost=2.279775898 test cost=2.278948682\n",
            "Epoch: 0067 train cost=2.279339030 test cost=2.278500741\n",
            "Epoch: 0068 train cost=2.278896486 test cost=2.278050893\n",
            "Epoch: 0069 train cost=2.278450201 test cost=2.277586519\n",
            "Epoch: 0070 train cost=2.277997427 test cost=2.277127512\n",
            "Epoch: 0071 train cost=2.277540361 test cost=2.276653729\n",
            "Epoch: 0072 train cost=2.277078350 test cost=2.276171335\n",
            "Epoch: 0073 train cost=2.276610667 test cost=2.275707240\n",
            "Epoch: 0074 train cost=2.276137298 test cost=2.275204777\n",
            "Epoch: 0075 train cost=2.275658953 test cost=2.274733351\n",
            "Epoch: 0076 train cost=2.275174474 test cost=2.274227627\n",
            "Epoch: 0077 train cost=2.274684853 test cost=2.273717456\n",
            "Epoch: 0078 train cost=2.274189373 test cost=2.273205914\n",
            "Epoch: 0079 train cost=2.273686140 test cost=2.272699835\n",
            "Epoch: 0080 train cost=2.273178508 test cost=2.272176253\n",
            "Epoch: 0081 train cost=2.272664129 test cost=2.271644532\n",
            "Epoch: 0082 train cost=2.272143000 test cost=2.271113611\n",
            "Epoch: 0083 train cost=2.271615998 test cost=2.270578122\n",
            "Epoch: 0084 train cost=2.271082051 test cost=2.270020327\n",
            "Epoch: 0085 train cost=2.270541198 test cost=2.269472645\n",
            "Epoch: 0086 train cost=2.269993820 test cost=2.268900779\n",
            "Epoch: 0087 train cost=2.269438528 test cost=2.268339914\n",
            "Epoch: 0088 train cost=2.268877673 test cost=2.267765911\n",
            "Epoch: 0089 train cost=2.268307866 test cost=2.267184824\n",
            "Epoch: 0090 train cost=2.267731835 test cost=2.266587436\n",
            "Epoch: 0091 train cost=2.267147981 test cost=2.265984981\n",
            "Epoch: 0092 train cost=2.266555378 test cost=2.265381098\n",
            "Epoch: 0093 train cost=2.265955834 test cost=2.264763596\n",
            "Epoch: 0094 train cost=2.265348358 test cost=2.264137203\n",
            "Epoch: 0095 train cost=2.264732162 test cost=2.263509328\n",
            "Epoch: 0096 train cost=2.264108045 test cost=2.262873700\n",
            "Epoch: 0097 train cost=2.263475292 test cost=2.262220856\n",
            "Epoch: 0098 train cost=2.262833993 test cost=2.261568221\n",
            "Epoch: 0099 train cost=2.262183753 test cost=2.260897731\n",
            "Epoch: 0100 train cost=2.261525156 test cost=2.260215246\n",
            "Accuracy: 0.3333\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAd4AAAEWCAYAAADIJfYaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCwleAFhI8ALAQj+Fxd2RJKnx/wAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LioBTnwBDXo3",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "866139af-1acf-4b01-b420-e045519165b3"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEMCAYAAADAqxFbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nnd9qWpzdbQAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FqFQ6XWVYJ6Q",
        "colab_type": "text"
      },
      "source": [
        "#Model 2: input (784) - ReLu(512) - ReLu(128) - sigmoid(output 10) "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Axi147ZWYJDh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Create model\n",
        "def multilayer_perceptron_relu(x, weights, biases):\n",
        "    # Use tf.matmul instead of \"*\" because tf.matmul can change it's dimensions on the fly (broadcast)\n",
        "    print( 'x:', x.get_shape(), 'W[h1]:', weights['h1'].get_shape(), 'b[h1]:', biases['b1'].get_shape())        \n",
        "    \n",
        "    # Hidden layer with ReLu activation\n",
        "    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) #(x*weights['h1']) + biases['b1']\n",
        "    layer_1 = tf.nn.relu(layer_1)\n",
        "    print( 'layer_1:', layer_1.get_shape(), 'W[h2]:', weights['h2'].get_shape(), 'b[h2]:', biases['b2'].get_shape())        \n",
        "    \n",
        "    # Hidden layer with ReLu activation\n",
        "    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) # (layer_1 * weights['h2']) + biases['b2'] \n",
        "    layer_2 = tf.nn.relu(layer_2)\n",
        "    print( 'layer_2:', layer_2.get_shape(), 'W[out]:', weights['out'].get_shape(), 'b3:', biases['out'].get_shape())        \n",
        "    \n",
        "    # Output layer with Sigmoid activation\n",
        "    out_layer = tf.matmul(layer_2, weights['out']) + biases['out'] # (layer_2 * weights['out']) + biases['out']    \n",
        "    out_layer = tf.nn.sigmoid(out_layer)\n",
        "    print('out_layer:',out_layer.get_shape())\n",
        "\n",
        "    return out_layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mnl9bLaHESux",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2224
        },
        "outputId": "770df04d-0240-4bf6-ad9d-5d2c2a64b63f"
      },
      "source": [
        "# Since we are using Relu activations in hiden layers we will be using weights that are initalized as weights_relu\n",
        "yrelu = multilayer_perceptron_relu(x, weights_relu, biases)\n",
        "\n",
        "# https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits\n",
        "cost_relu = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = yrelu, labels = y_))\n",
        "# https://github.com/amitmac/Question-Answering/issues/2\n",
        "# there are many optimizers available: https://www.tensorflow.org/versions/r1.2/api_guides/python/train#Optimizers \n",
        "optimizer_relu_adam = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost_relu)\n",
        "optimizer_relu_sgdc = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost_relu)\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    \n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use AdamOptimizer\n",
        "            _, c, w = sess.run([optimizer_relu_adam, cost_relu, weights_relu], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_relu, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-ReLu(512)-ReLu(128)-sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax,np.arange(1.3, 1.8, step=0.04), \"input-ReLu(512)-ReLu(128)-sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(yrelu,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x: (?, 784) W[h1]: (784, 512) b[h1]: (512,)\n",
            "layer_1: (?, 512) W[h2]: (512, 128) b[h2]: (128,)\n",
            "layer_2: (?, 128) W[out]: (128, 10) b3: (10,)\n",
            "out_layer: (?, 10)\n",
            "Epoch: 0001 train cost=1.558863960 test cost=1.553809897\n",
            "Epoch: 0002 train cost=1.500869568 test cost=1.501984452\n",
            "Epoch: 0003 train cost=1.488610380 test cost=1.492479212\n",
            "Epoch: 0004 train cost=1.482727066 test cost=1.489153654\n",
            "Epoch: 0005 train cost=1.478434366 test cost=1.486031231\n",
            "Epoch: 0006 train cost=1.475590462 test cost=1.484320518\n",
            "Epoch: 0007 train cost=1.473480541 test cost=1.482918699\n",
            "Epoch: 0008 train cost=1.472103152 test cost=1.482520484\n",
            "Epoch: 0009 train cost=1.470486153 test cost=1.481997091\n",
            "Epoch: 0010 train cost=1.469727064 test cost=1.481558785\n",
            "Epoch: 0011 train cost=1.468763274 test cost=1.480968859\n",
            "Epoch: 0012 train cost=1.468208797 test cost=1.481250223\n",
            "Epoch: 0013 train cost=1.468463499 test cost=1.480946891\n",
            "Epoch: 0014 train cost=1.467842502 test cost=1.480608781\n",
            "Epoch: 0015 train cost=1.467086066 test cost=1.480197428\n",
            "Epoch: 0016 train cost=1.466656428 test cost=1.480296041\n",
            "Epoch: 0017 train cost=1.465965633 test cost=1.480015448\n",
            "Epoch: 0018 train cost=1.465989232 test cost=1.479685502\n",
            "Epoch: 0019 train cost=1.466056588 test cost=1.480470149\n",
            "Epoch: 0020 train cost=1.465800858 test cost=1.480406627\n",
            "Epoch: 0021 train cost=1.465139879 test cost=1.479596285\n",
            "Epoch: 0022 train cost=1.465369952 test cost=1.479517240\n",
            "Epoch: 0023 train cost=1.465357349 test cost=1.479149662\n",
            "Epoch: 0024 train cost=1.465247758 test cost=1.478957816\n",
            "Epoch: 0025 train cost=1.465198136 test cost=1.479512174\n",
            "Epoch: 0026 train cost=1.464558409 test cost=1.479033244\n",
            "Epoch: 0027 train cost=1.464449206 test cost=1.479749947\n",
            "Epoch: 0028 train cost=1.464598196 test cost=1.479335078\n",
            "Epoch: 0029 train cost=1.464560271 test cost=1.478832364\n",
            "Epoch: 0030 train cost=1.464906664 test cost=1.479765068\n",
            "Epoch: 0031 train cost=1.464397625 test cost=1.478966192\n",
            "Epoch: 0032 train cost=1.464484556 test cost=1.479443267\n",
            "Epoch: 0033 train cost=1.464194885 test cost=1.478376426\n",
            "Epoch: 0034 train cost=1.463769236 test cost=1.478565987\n",
            "Epoch: 0035 train cost=1.463827819 test cost=1.479032257\n",
            "Epoch: 0036 train cost=1.464321488 test cost=1.479465319\n",
            "Epoch: 0037 train cost=1.464392973 test cost=1.478818833\n",
            "Epoch: 0038 train cost=1.464101514 test cost=1.479069341\n",
            "Epoch: 0039 train cost=1.464351427 test cost=1.479261617\n",
            "Epoch: 0040 train cost=1.463434665 test cost=1.478782720\n",
            "Epoch: 0041 train cost=1.464059338 test cost=1.478533023\n",
            "Epoch: 0042 train cost=1.464080035 test cost=1.478328096\n",
            "Epoch: 0043 train cost=1.463975060 test cost=1.478034090\n",
            "Epoch: 0044 train cost=1.463707302 test cost=1.478643742\n",
            "Epoch: 0045 train cost=1.463918767 test cost=1.478996321\n",
            "Epoch: 0046 train cost=1.464037966 test cost=1.479502049\n",
            "Epoch: 0047 train cost=1.463564400 test cost=1.478774913\n",
            "Epoch: 0048 train cost=1.463892798 test cost=1.479036026\n",
            "Epoch: 0049 train cost=1.463975043 test cost=1.478849379\n",
            "Epoch: 0050 train cost=1.463627027 test cost=1.478102100\n",
            "Epoch: 0051 train cost=1.463829713 test cost=1.478803757\n",
            "Epoch: 0052 train cost=1.463960039 test cost=1.479185219\n",
            "Epoch: 0053 train cost=1.463917185 test cost=1.478309902\n",
            "Epoch: 0054 train cost=1.463755745 test cost=1.478627446\n",
            "Epoch: 0055 train cost=1.463989246 test cost=1.478895878\n",
            "Epoch: 0056 train cost=1.463441450 test cost=1.478186780\n",
            "Epoch: 0057 train cost=1.463480893 test cost=1.478294485\n",
            "Epoch: 0058 train cost=1.463425806 test cost=1.478487250\n",
            "Epoch: 0059 train cost=1.463650924 test cost=1.477561233\n",
            "Epoch: 0060 train cost=1.463438427 test cost=1.478694949\n",
            "Epoch: 0061 train cost=1.463538481 test cost=1.478432217\n",
            "Epoch: 0062 train cost=1.463763734 test cost=1.478327941\n",
            "Epoch: 0063 train cost=1.463436301 test cost=1.477445348\n",
            "Epoch: 0064 train cost=1.463418414 test cost=1.477949221\n",
            "Epoch: 0065 train cost=1.463492493 test cost=1.477386128\n",
            "Epoch: 0066 train cost=1.463452673 test cost=1.477535840\n",
            "Epoch: 0067 train cost=1.463452720 test cost=1.477597484\n",
            "Epoch: 0068 train cost=1.463264907 test cost=1.477486010\n",
            "Epoch: 0069 train cost=1.463683417 test cost=1.478171059\n",
            "Epoch: 0070 train cost=1.463707072 test cost=1.478055995\n",
            "Epoch: 0071 train cost=1.463676369 test cost=1.477555312\n",
            "Epoch: 0072 train cost=1.463745266 test cost=1.477359731\n",
            "Epoch: 0073 train cost=1.463477546 test cost=1.478063452\n",
            "Epoch: 0074 train cost=1.463263168 test cost=1.477570219\n",
            "Epoch: 0075 train cost=1.463014812 test cost=1.476893024\n",
            "Epoch: 0076 train cost=1.464126511 test cost=1.478066534\n",
            "Epoch: 0077 train cost=1.463602071 test cost=1.477420642\n",
            "Epoch: 0078 train cost=1.463259823 test cost=1.477705346\n",
            "Epoch: 0079 train cost=1.463384225 test cost=1.477744463\n",
            "Epoch: 0080 train cost=1.463387481 test cost=1.478071987\n",
            "Epoch: 0081 train cost=1.463595021 test cost=1.478020390\n",
            "Epoch: 0082 train cost=1.463454898 test cost=1.477327898\n",
            "Epoch: 0083 train cost=1.463283242 test cost=1.477973552\n",
            "Epoch: 0084 train cost=1.463196347 test cost=1.477012328\n",
            "Epoch: 0085 train cost=1.462736135 test cost=1.477902700\n",
            "Epoch: 0086 train cost=1.463323203 test cost=1.477934962\n",
            "Epoch: 0087 train cost=1.463846704 test cost=1.478496767\n",
            "Epoch: 0088 train cost=1.463812223 test cost=1.478132971\n",
            "Epoch: 0089 train cost=1.463478042 test cost=1.477640873\n",
            "Epoch: 0090 train cost=1.463622350 test cost=1.477558313\n",
            "Epoch: 0091 train cost=1.463577070 test cost=1.477025384\n",
            "Epoch: 0092 train cost=1.463296953 test cost=1.477881937\n",
            "Epoch: 0093 train cost=1.463218552 test cost=1.477153061\n",
            "Epoch: 0094 train cost=1.462860535 test cost=1.477478754\n",
            "Epoch: 0095 train cost=1.463111204 test cost=1.478222600\n",
            "Epoch: 0096 train cost=1.463345155 test cost=1.478163169\n",
            "Epoch: 0097 train cost=1.464227972 test cost=1.478730225\n",
            "Epoch: 0098 train cost=1.463464970 test cost=1.478189470\n",
            "Epoch: 0099 train cost=1.463691602 test cost=1.477813894\n",
            "Epoch: 0100 train cost=1.463370332 test cost=1.476827080\n",
            "Accuracy: 0.9832\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZYAAAEWCAYAAABFSLFOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n5cBiZmYV5cBiZmYV5cBiZmYV9f8B3IUQGwlgIZYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pbb81KtqbKwO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "8914f6f1-4cae-4b02-a56a-b64d7913dcdd"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEMCAYAAADNtWEcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gamuhAb1fewy",
        "colab_type": "text"
      },
      "source": [
        "**ReLu(512)-ReLu(128)-sigmoid(output) - GradientDescentOptimizer**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vntDnwkwfZbu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1232
        },
        "outputId": "3fb64c11-160a-44d9-a8d2-e1e1a1e4da8e"
      },
      "source": [
        "# We can now launch the model in an InteractiveSession\n",
        "\n",
        "# We first have to create an operation to initialize the variables we created:\n",
        "\n",
        "\n",
        "# Make sure to initialize variables after AdamOptimizer\n",
        "\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use GradientDescentOptimizer\n",
        "            _, c, w = sess.run([optimizer_relu_sgdc, cost_relu, weights_relu], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_relu, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.5, 2.4, step=0.05), \"input-ReLu(512)-ReLu(128)-sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.5, 2.4, step=0.05), \"input-ReLu(512)-ReLu(128)-sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(yrelu,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 0001 train cost=2.321509665 test cost=2.316798456\n",
            "Epoch: 0002 train cost=2.227539322 test cost=2.219945126\n",
            "Epoch: 0003 train cost=2.144788755 test cost=2.135815686\n",
            "Epoch: 0004 train cost=2.067583695 test cost=2.058638025\n",
            "Epoch: 0005 train cost=1.989005513 test cost=1.979689266\n",
            "Epoch: 0006 train cost=1.934081882 test cost=1.924469748\n",
            "Epoch: 0007 train cost=1.895377707 test cost=1.885935935\n",
            "Epoch: 0008 train cost=1.866585492 test cost=1.857340544\n",
            "Epoch: 0009 train cost=1.844205585 test cost=1.835351400\n",
            "Epoch: 0010 train cost=1.826191737 test cost=1.817553650\n",
            "Epoch: 0011 train cost=1.811140100 test cost=1.802762371\n",
            "Epoch: 0012 train cost=1.797553562 test cost=1.789487024\n",
            "Epoch: 0013 train cost=1.781139457 test cost=1.773214935\n",
            "Epoch: 0014 train cost=1.765221260 test cost=1.757221149\n",
            "Epoch: 0015 train cost=1.753694408 test cost=1.745608294\n",
            "Epoch: 0016 train cost=1.743808135 test cost=1.735641602\n",
            "Epoch: 0017 train cost=1.735142255 test cost=1.726996134\n",
            "Epoch: 0018 train cost=1.727488239 test cost=1.719296534\n",
            "Epoch: 0019 train cost=1.720611890 test cost=1.712434181\n",
            "Epoch: 0020 train cost=1.714382583 test cost=1.706266644\n",
            "Epoch: 0021 train cost=1.708686155 test cost=1.700572264\n",
            "Epoch: 0022 train cost=1.703432715 test cost=1.695370186\n",
            "Epoch: 0023 train cost=1.698567147 test cost=1.690548376\n",
            "Epoch: 0024 train cost=1.694039845 test cost=1.686053524\n",
            "Epoch: 0025 train cost=1.689789891 test cost=1.681857559\n",
            "Epoch: 0026 train cost=1.685807971 test cost=1.677937715\n",
            "Epoch: 0027 train cost=1.682051042 test cost=1.674233088\n",
            "Epoch: 0028 train cost=1.678499075 test cost=1.670756419\n",
            "Epoch: 0029 train cost=1.675144906 test cost=1.667479436\n",
            "Epoch: 0030 train cost=1.671954563 test cost=1.664333004\n",
            "Epoch: 0031 train cost=1.668943664 test cost=1.661376875\n",
            "Epoch: 0032 train cost=1.666065937 test cost=1.658529018\n",
            "Epoch: 0033 train cost=1.663321029 test cost=1.655828080\n",
            "Epoch: 0034 train cost=1.660692391 test cost=1.653257967\n",
            "Epoch: 0035 train cost=1.658191235 test cost=1.650779156\n",
            "Epoch: 0036 train cost=1.655786594 test cost=1.648438390\n",
            "Epoch: 0037 train cost=1.653489083 test cost=1.646150701\n",
            "Epoch: 0038 train cost=1.651288229 test cost=1.643990705\n",
            "Epoch: 0039 train cost=1.649164474 test cost=1.641890808\n",
            "Epoch: 0040 train cost=1.647131652 test cost=1.639895172\n",
            "Epoch: 0041 train cost=1.645162777 test cost=1.637965748\n",
            "Epoch: 0042 train cost=1.643275636 test cost=1.636122550\n",
            "Epoch: 0043 train cost=1.641451280 test cost=1.634297085\n",
            "Epoch: 0044 train cost=1.639690609 test cost=1.632590287\n",
            "Epoch: 0045 train cost=1.637981496 test cost=1.630902398\n",
            "Epoch: 0046 train cost=1.636348970 test cost=1.629294222\n",
            "Epoch: 0047 train cost=1.634753797 test cost=1.627740584\n",
            "Epoch: 0048 train cost=1.633223719 test cost=1.626215783\n",
            "Epoch: 0049 train cost=1.631729781 test cost=1.624750015\n",
            "Epoch: 0050 train cost=1.630292597 test cost=1.623368808\n",
            "Accuracy: 0.8805\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbsAAAEWCAYAAAD/6zkuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3XecVNX9//HXm2UBpRcRpdsF1BVX\n1Ig1xBIlaqIm9qCG+ItGjV2/9hRLjFGjiTGKxpIQjZpgLNgxioYmoFIUEQQEaSqgKOzu5/fHOQuz\nw+xO2dmdLZ/n4zGP3b31zN0787nn3HM/R2aGc84515S1KHQBnHPOubrmwc4551yT58HOOedck+fB\nzjnnXJPnwc4551yT58HOOedck5c22El6T9KB9VCWBkvSoZL+VQ/72VXS+LreT64kbSFplqTN6mFf\nj0s6PMd1T5L0fL7LVNv9SnpV0pkJf9fb8awtSVtKmimpdQMoSz9JJqll/PtZSacVulzNjaS7JV2V\n47pXSLo332WqkZkV/AU8APwqzTLXAuuBNcDnwHhgnyz2MQ8YlmP5JgF7J21rbSzLGuD5hHmDgLHA\n8nB4q2ynNXAfMB9YDUwFDk9a5hlgeA1lORCoiPtdDcwGRuTzWNew7u+AyxL+Pj7+H74CXk1adgfg\n38AyYGU8JjsmzBfwK2AR8AXwKjAwYf4QYHKhz818vuJ7PLO641nLbed8fme6PeCPwM/TrNcKuDqe\nl1/G/++zwCF5LFs/wICWef7//Bh4PWnaA8C6+FlbDbwL3AB0LPT5lM2xAXoBjwAr4v9lAnBkbY5N\nY3s1tmbMf5hZO6Ab8ArwWF3vUNKehBP7raRZw82sXXwdkjB9PfAocEaKzbUEFgAHAB2BK4FHJfVL\nWOYR4KdpivVJPA4dgF8Af5G0Y4ZvKSfxiv404OGEySuB24AbU6zSCRgD7AhsSfhw/Tth/nHA6cB+\nQBfgTeChyplmNgHoIKk0f++i4ajmeDZ0mZyb/wSOAk4FOgP9gduBI1ItXFk7a+BuNrP2wBbACGBv\n4A1JbQtbrMxI6gK8TgjaAwnfn78H/ibp2EKWLZ/SnksZRPR5wDBCzepR4EHCFc57QGnScpcDM4DP\ngPuBNjVcMRmwHTCSECDWEWorT1VTjmuBhxP+HhC3sUXCtCMJtaXKmt+uye8jkyuWyrLF368G7k11\nTNIct+1IqtlVs9x04AcJf/ck1BpbV7P8gcDCpGlLgeMS/t4JeIEQjGYDxyfMe4AUNTtSXBGSUBMB\n9gfmVFOmM0mq2aVYpkvcftf496XAownzBwJfJ63zF+CaGrb5Y2BuPB8/Ak5K9T8FDonH4QtC7WRc\nwvv6MfAG4cP/edzet+L0BfHYnpawrY6Ez8AyQg39SqBFNfv9DjAr7vfOpP1ucjyBrQkXCCuBOcBP\nqvu/JZ4HhIuECja2NlyS8P8cCXwCLAYuynV7cXpLQi2+bzX/j2FxnV4ZfKdcSjj3v4nbvQz4MP4v\nZwDHJCxfBNxCaC2ZC5xNwrnKpjXm04GZhO+hsYnljeudBXwQ/993EVoZdga+Bsrje/68us8L0D4e\nz3PS7TNu+/eE82gV8A4wKM7bjFC7nx/PkdeBzeK8vQnfYZ8D04ADkz6XvySct6uB54Fucd7H8T1W\ntjrtE5d9l3ieJmzn0rhvJRybc+MxXg78lnCrK+2xIZ4/hHNvaTw+RwPfBd4nnNNXJOz7WuL3OeGz\nsSbhVQZcm/CZeJzwefsIODdpG/8kXDCuSjwHUr2yrdl9DxjNxqv2O5PmnwQcCmxLaMa6Mt0Gzewe\nwhXjzRZqScPTrSOpFeHKcQXh5ELS7sAowpVnV+DPwJg83GPYhfBFmewRScskPS9pt1w2LGlLwnF6\nr3KamS0iBP+0NTVJLSR9j3ClNidOa0sIdH8DugM/Av4oaUAuZUxQ3XHI1P7AEjNbEf8eDWwraQdJ\nxYRaznNJ68wEUh7b+D7vIDQDtycEqKkplutG+EBcTjgvZsdlE+1F+OLtSjhuo4E9CRcsJwN3SmoX\nl/0DIeBtQ6ihn0q42k+13ycIn4FuhC/yfRMWSXU8RxO+MLYGjgV+I+ngVO8/kZmdQviSq2xtuDlh\n9kHA9oSAf6mkYbluz8zKCOdZdef7MOB/ZrYw3T6AEwi1vU5xux8SavkdgeuAhyVtFZf9CeFCdneg\nlHBsUpJ0FHAF8H1CTey/wN+TFjuS8P/dldAUf6iZzSQEwTfje+5U3T7MbDXhM7ZfBvs8hHDu7xDf\n2/GE7y0IAXwPwvnYhRAoKiT1BJ4mNPN3AS4CHpe0RUIxTiScd90JTccXxen7x5+d4vt4k3DR9biZ\nVSS9lUeBPrFslY4hHOPBhBr66Vkcmx5AG8IF+9WEi9WT43vcD7hKUv/klczsnLjddsBQwnf6vyW1\nAJ4iBPuewLeB8yUdmrD6UYTPdydCHKlWtsHudTN7xszKCVd/ySf9nWa2wMxWAr8mnND5dLykzwlX\njz8Bjo0fFAhXsH82s/+ZWbmZ/ZVw1bh3LffZiXD1lOgkwpVzX0Jz6lhJ1X44Uolf8I8AfzWzWUmz\nV8f9VmfrhOPwJHCBmb0d5x0JzDOz+82sLE5/nNBsWBupjkNGJPUiXEFfkDB5MeFKdjbhfRxHaJJN\nlO44VACDJG1mZovN7L0Uy3wXeM/Mnojnyh3AkqRlPorHqxz4B9AbuN7MvjGz5wmtDttJKiJcPFxu\nZqvNbB7hyvyUGvb7TzNbT2juTdxvleMpqTchGF5qZl+b2VTgXkIwrY3rzOxLM3uH0NpS289kTf+T\nbiS8R0ldJH0u6QtJXycte0f8rlgLYGaPmdknZlZhZv8g1LyGxGWPB25L+G65oYbynQXcYGYz4//7\nN0CJpL4Jy9xoZp+b2ceEz29JZm+9ik8IgSjdPtcTaoI7EWpQM81scfwiPx04z8wWxe+s8Wb2DSFA\nPBO/ayvM7AVCv4HvJuz/fjN7Px6/R9O8h26Ez1uyxQnzK91kZivjsbmN7M6X9cCv4/k+Om739vhZ\neY9QY6+2YhCD+b8I94XfJlyQbGFm15vZOjObSwigP0pY7U0z+1c8TmtrKly2wS7xw/oV0CapnXRB\nwu/zCVeoWYu92tbE17MJsx6NVxVbEqrleyTM6wtcGD9cn8dg0DvXMiT4jHCybmBmb5jZWjP7ysxu\nIDQ17JfpBuOJ/hDhS/ScFIu0Bz6X1CfhOKxJmP9JPA4dCF/eiVf/fYG9ko7DSYSrrtrY5DhkIp7A\nzwN/NLPEK+yrCSdzb8LV4HXAy5I2T1imPeHYVvb8qjwWV5jZl8APCV80iyU9LWmnFEXYmoTz0kL7\nR3LN49OE3yu/fJOnVd4rLiac25XmE646M9lv4ucj+XhuDayMtYZ0285GXj6TCRL/J2sSXn0INZbK\n2hjxS7MT4XOa3MKSWC4knSppasI5O4iNX8Jbs+n7qE5f4PaE7awkNCUmHsfk77F2ZK9n3HaN+zSz\nlwktYHcBSyXdI6lDfG9tCDXaVO/huKTP8FASjm2W72F50rqVtkqYX6k258uKeMEI8XPEpp+tlOWM\nF///BP5mZqPj5L7EC/uE43AF4fs/VXlrlO8OKr0Tfu9DuPqB0Ptnw5eYpOQvXqvyh9kjtrHzxybd\nz81sOaEmd21CU8cCwlVFp4TX5klfsKmkK9t0qlbzUzHCyZ2WJBF6ZG5JuFe3Pml+T0KzxGwz+zjh\nOGxyksSrwEuBXSQdHScvAMYlHYd2Zvb/0hTty/gzMdgkHotMjkMVkjoTAt0YM/t10uwSQoejhbEG\n+gChQ0Nic+vOhCYMzOyshGPxmzhtrJl9h/ChnUW46ku2mNATrbJMSvw7S8sJV6+JtYQ+hB6Hqfa7\n4fMQ95v4+Ug+np8AXSQlBsDEbVc5T9n04sVILaPPZCbbixe227Hxf9Iu4fUx8BKwZ6zJp7Nh+7EG\n9BfChV/XGCDfZeNnqsqxjO+jOguAnyad/5uZWSaP9FR3DKuITdrDCM2VafdpZneY2R6Ec3sH4GLC\nufQ14ZZPqvfwUNL22ppZqo5gmbyHF4Hvx4vsRMfHfb2fMK268yWjY1MLfyDcd0u89bWA0OqSeBza\nm1liDTfjcuU72J0tqVfs/fN/hCYhCB+OgZJKJLUh3FhM9CnhHkjGzGw24UbwJXHSX4CzJO2loK2k\nI5K+PIoltUl4tcygbM8Q7s0AEGtb+0pqFbdxMeEq7Y04X3E7reLfbVT1vuGfCF/iw6updh8AvBwD\nWSbHYR2hKe3qOOk/wA6STpFUHF97Sto5YbWipOPQysyWEb5YT5ZUJOl0qn4QJwCdYjCuPBZF8b22\nBFrEbRXHeR0I/583zOyyFEWfSLh63VLh3uMphFrTnKRj8WyKdSuf+zpK4d7dN4Qb28n3JCDc+9hF\n0tHx/302OdZy41Xro8CvJbWPX9IXkLpH5dOE8+r7cb/nJu23yvE0swWEDgk3xOO4K6FHb+W2pwLf\nVWga7AGcn7S/6j5DV0naXNJAwj2eys9kLtsbQmgiT1mzstDk+wrwr/g5bBXPh3S3EtoSvrSWAUga\nQajZVXoUODd+t3QmdGapzt3A5fH9IqmjpEyb8D8Fein0CdiEpNaS9iA0tVV2wqtxn/Gzt1c8Dl8S\nAlyFhftno4BbJW0dP0v7xO+Kh4HhCs/3Vn5WD8zwImIZ4XOQ+L/7PeF+4X2SesTtnUD4jr44tjpU\nulhSZ4Vm9fPYeL7UeGxqQ9JPCZ/1k6zqfcUJwGpJl0raLB6LQQo95LOW72D3N8KV/FxC9fxXAGb2\nPnA94QrjA8K9mkT3AQNiVTWbh7d/C4yU1N3MJhHu491JOBHnEHrHJXqGUJWufF2brmxmNgX4QtJe\ncVJ7QsD6jBAcDiN0kqi86dw3brvy/tFaYkeE+OX4U0KtZok2NgGdlLDLkwgfnmyMAvpIGh6bwQ4h\ntGt/QmjuuImqzUiXJR2Hl+P0nxCuOlcQekduuBqOQfUBwv2ESqfE9f9EaMZdy8ba1TGEZsoR2rS5\ni1imaWzsPfsLQk23solsT2CNhUcQUmlBCDSfEJqNDgA2qb3GVoDjgJvj+xpAuP+R0cVECj8nfGnN\nJZwrfyMc/+r2e2Pc7/bEC6I4P9XxPIFwL/gTwr3Ya8zsxTjvIcLxmkf4jP2Dqm4AroyfoYsSpo8j\nfBZeAm6JASnX7WVybh5DuOB6mPB//YiNHddSMrMZhAu2NwlfqruQcKwI59TYWN4phI4/1W3rScK5\nNVrSKkINMdPkBC8TPrdLJCU27V0iaTXh//ggMBn4VmxKT7fPDrH8nxGaBVcQvrcgdCp5h3DhtzJu\no0W88Kns9LKMUMO5mAy+r83sK0J/iTfi/27v+N00lNBsOiOW4QLgFAv3RxP9O76/qYQLtvvSHJt8\nOIEQnD9R1VsV5YQ+CCWE82g54T52x1x2oqpBPXeS5hG6fr6YbtnGRtIhwM/M7Oi0C9duP7sSOtns\nU5f7yZXC/bf/Arunuxmch309DtxnZs/kebstCPfsTjKzV/K57RzKUmfHU+HZzY+AYtvYias22+tO\nCJy7m1lyZxPXBEgyYHszm5N24UbIg51r8hS6Kv+PUPO8mNCUuU1dB+xCynewc01fUw92jS2DinO5\n2IfQrL4cGA4c3ZQDnXNuU3mr2TnnnHMNVb3V7CT1lvSKpBkKIymcl2KZoyRNV3jeZpKkoQnzyuP0\nqZLG1Fe5nXPONX71VrNTeB5uKzObovA4wGRCc9KMhGXaAV+amcXOGo+a2U5x3ppUz5pVp1u3btav\nX7/8vgnnnGviJk+evNzMtki/ZONSbxnHzWwxMT2Nma2WNJOQhWBGwjKJWUIqn73JSb9+/Zg0aVKu\nqzvnXLMkqaYMNY1WQTqoxJ5iuxN6yCXPO0bSLMIzHqcnzGoTmzbf0sZsIcnrjozLTFq2bFkdlNw5\n51xjVO/BLjZVPg6cb2arkueb2ZOx6fJowtAUlfqaWSkh2/dtkjZJs2Nm95hZqZmVbrFFk6uFO+ec\ny1G9BruYMudx4BEzqzYLAoCZvQZsozBUSuXQN1jIfP0qoWbonHPOpVVv9+ykDQmQZ5rZrdUssx3w\nYeygMpiQ4mpFzIf3lZl9E4PfvoT0T865ZmD9+vUsXLiQr7/25C350qZNG3r16kVxcXGhi1Iv6i3Y\nEQLUKcA7kioH2byCmMHczO4GfgCcKmk9IdvFD2Pg2xn4s6QKQm30xsRenM65pm3hwoW0b9+efv36\nEa6bXW2YGStWrGDhwoX077/JeKpNUn32xnydNMPgmNlNhGSoydPHE5LDOueaoa+//toDXR5JomvX\nrjSnjnyeLixJURFI4eWcazg80OVXczue9dmM2SicV3ElcyllBV2A/QtdHOecc3ngNbskt7a4gSf4\nPkeTzbB6zrmmbMWKFZSUlFBSUkKPHj3o2bPnhr/XrVuX0TZGjBjB7NmzM97nvffey/nnJ4+p63Ll\nNbtk5eV8ol6UMhkJPE+2c65r165MnRr61V177bW0a9eOiy66qMoyZoaZ0aJF6jrE/fffn3K6qx+N\nKRH0aZI+iK/T6rKsq2nH3rxFDxbW5W6cc43cnDlzGDBgACeddBIDBw5k8eLFjBw5ktLSUgYOHMj1\n11+/YdmhQ4cydepUysrK6NSpE5dddhm77bYb++yzD0uXLs14nw8//DC77LILgwYN4oorrgCgrKyM\nU045ZcP0O+64A4Df//73DBgwgF133ZWTTz45v2++kanPml0ZcGFiImhJLyQ9QvASMCYxETSwk6Qu\nwDVAKSFf5mRJY8zss7ooaF/m0Zp1HMJLQJ3GVedcls4/H6ZOTb9cNkpK4Lbbclt31qxZPPjgg5SW\nlgJw44030qVLF8rKyjjooIM49thjGTBgQJV1vvjiCw444ABuvPFGLrjgAkaNGsVll12Wdl8LFy7k\nyiuvZNKkSXTs2JFhw4bxn//8hy222ILly5fzzjvvAPD5558DcPPNNzN//nxatWq1YVpzVW81OzNb\nbGZT4u+rgcpE0InLrLGNwzAkJoI+FHjBzFbGAPcCcFhdlXXz4grKacGuTK+rXTjnmohtt912Q6AD\n+Pvf/87gwYMZPHgwM2fOZMaMTR8J3myzzTj88MMB2GOPPZg3b15G+/rf//7HwQcfTLdu3SguLubE\nE0/ktddeY7vttmP27Nmce+65jB07lo4dOwIwcOBATj75ZB555JFm8/B4dQpyzy5dImjgBqA7cESc\n3BNYkLDYQpICZVx3JDASoE+fPrkXcN06lqgnQ5jg9+2ca2ByrYHVlbZt2274/YMPPuD2229nwoQJ\ndOrUiZNPPjll1pdWrVpt+L2oqIiysrJalaFr165Mnz6dZ599lrvuuovHH3+ce+65h7FjxzJu3DjG\njBnDb37zG6ZPn05RUVGt9tVYNaZE0GnlMxH017RmL/7Hliyq1Xacc83HqlWraN++PR06dGDx4sWM\nHTs2r9vfa6+9eOWVV1ixYgVlZWWMHj2aAw44gGXLlmFmHHfccVx//fVMmTKF8vJyFi5cyMEHH8zN\nN9/M8uXL+eqrr/JansakXmt22SaCllSZCHoRcGDC7F6EZNB1pg/zKaaCQ3gRv2/nnMvE4MGDGTBg\nADvttBN9+/Zl3333rdX27rvvPv75z39u+HvSpEn88pe/5MADD8TMGD58OEcccQRTpkzhjDPOwMyQ\nxE033URZWRknnngiq1evpqKigosuuoj27dvX9i02WvU5UrmAvwIrzSzlwyMpEkE/RQhsnQkjmw+O\ni04B9jCzldXtr7S01Go1eGu7dpR/uZbf8wsOnnwLgwenX8U5VzdmzpzJzjvvXOhiNDmpjqukyXE4\ntSalUSSCBlZK+iUwMa53fU2BLi/WrGGJtuZbjGePPfy+nXPONWaNIhF0nDcKGFUHRatWBS3Yk4ls\nzQKgd33u2jnnXB55urAabM0iiinjIF4pdFGcc87Vgge7GhR17UoZRZQwjWuuKXRpnHPO5cqDXU2W\nL2cZWzCUN0jI+uOcc66RaWi5MU+KuTHfkTRe0m4J8+bF6VMl1aKbZXaKKKOUSfG+nXPOucaoPmt2\nlbkxBwB7A2dLGpC0zEfAAWa2C+GB8nuS5h9kZiX12S22O8tpSbnft3OuGcvHED8Ao0aNYsmSJSnn\nnXzyyfzrXz60WF2pz96Yi4HF8ffVkipzY85IWGZ8wipvEZ6xK6wtt6Ts0+XszjQOPhhefrnQBXLO\n1bdMhvjJxKhRoxg8eDA9evTIdxFdGgW5Z1dTbswEZwDPJvxtwPOSJsccmKm2OzIODTRp2bJl+Sns\nkiUspyv78xqveOXOOZfkr3/9K0OGDKGkpISf/exnVFRUpBxy5x//+AdTp07lhz/8YcY1woqKCi64\n4AIGDRrELrvssiGbyqJFixg6dCglJSUMGjSI8ePHVzvMjwvqPRF0utyYcZmDCMFuaMLkoWa2SFJ3\n4AVJs8zstcT1zOweYtNnaWlp3h4Db8PXDGYKPfmY+Ay8c65QGtAYP++++y5PPvkk48ePp2XLlowc\nOZLRo0ez7bbbbjLkTqdOnfjDH/7AnXfeSUlJSUbbf+yxx5g5cybTpk1j2bJl7Lnnnuy///48/PDD\nDB8+nEsvvZTy8nLWrl3L5MmTUw7z44J6rdllkhszjmN3L3CUma2onG5mi+LPpcCTwJC6L3HQiVUU\nUcEBvJZ+Yedcs/Hiiy8yceJESktLKSkpYdy4cXz44YfVDrmTrddff50TTjiBoqIievTowdChQ5k0\naRJ77rkn9957L9dddx3vvvsu7dq1y9s+m6p6q9nF3Jj3ATPN7NZqlukDPAGcYmbvJ0xvC7SI9/ra\nAocA9fcwwE47sX7WHPZgMt26nczy5fW2Z+dcsgY0xo+Zcfrpp/PLX246QEuqIXfy5eCDD+bVV1/l\n6aef5tRTT+WSSy7hpJNOqtN9Nnb1WbOrzI15cHx8YKqk70o6S9JZcZmrga7AH5MeMdgSeF3SNGAC\n8LSZPVdvJZ85k8/ozIGMY8WK9Is755qHYcOG8eijj7I8XgGvWLGCjz/+OOWQOwDt27dn9erVGW9/\nv/32Y/To0VRUVPDpp5/yxhtvUFpayvz58+nRowcjR45kxIgRvP3229Xu0wUNLTfmmcCZKabPBXbb\ndI3605Y1lDCV3swH+hayKM65BmKXXXbhmmuuYdiwYVRUVFBcXMzdd99NUVHRJkPuAIwYMYIzzzyT\nzTbbjAkTJlQZxBXgzDPP5JxzzgGgf//+jBs3jrfeeotdd90VSdx66610796dUaNGceutt1JcXEz7\n9u156KGHWLBgQcp9uqDehvipb7Ue4ieZQpw+kUf4m52Yv+0659LyIX7qRnMa4sfThWVq8GDW05Ih\nTKiMe8455xoJD3aZmjyZtWzGoYwtdEmcc85lyYNdFtqxmp2ZxSCmFboozjU7TfWWS6E0t+PZmBJB\nHyZptqQ5ki6rr3InqjxYB/EqW2xRiBI41zy1adOGFStWNLsv6LpiZqxYsYI2bdoUuij1pj4zqFQm\ngp4iqT0wWdILZjYjYZnKRNCfSTqckA1lL0lFwF3Ad4CFwERJY5LWrXtmfKm2HMZY/rB8k1jtnKsj\nvXr1YuHCheQtDaCjTZs29OpV+PTD9aWxJIIeAsyJjyAgaTRwVOK69aUV33AQr9CRz4DO9b1755ql\n4uJi+vfvX+hiuEassSSC7glVBpRbGKclbzf/iaCTFFPOZnzNQbzC22/XyS6cc87lWb0HuywTQV+a\nzbbN7B4zKzWz0i3q6qZaaSnracl3eJHBg+tmF8455/Ir62An6WZJHSQVS3pJ0jJJJ2e4bq6JoBcB\nvRMW6xWn1b+JE1lHMd/lGaC8IEVwzjmXnVxqdofEGtmRwDxgO+DidCvVJhE0MBHYXlJ/Sa2AHwFj\ncih7XrRlLf2YTymee8455xqDXDqoVK5zBPCYmX2hzFKKVCaCfkdS5WBUVxAHiDOzu6maCBqgLDZL\nlkk6BxgLFAGjzOy9HMqeVwfyCtKeeG9o55xr2HIJdv+RNAtYC/w/SVsAX6dbqTaJoOO8Z4Bnsi9u\nHYiPIBzBM9zCJYUujXPOuTSybsY0s8uAbwGlZrYe+JLwGECzsjlfMZTX6canhS6Kc865NHLpoHIc\nsN7MyiVdCTwMbJ33kjVwAlpSzrd5hWHDCl0a55xzNcmlg8pV8aHwocAwQqeTP+W3WI3AT3/KOoo5\nlLG89FKhC+Occ64muQS7yv72RwD3mNnTQKsalm+a7r4bAw7nWURZoUvjnHOuBrkEu0WS/gz8EHhG\nUutMtpNhIuidJL0p6RtJFyXNmxcTRE+VlMdRWXPXmvX04FP24a1CF8U551wNcumNeTxwGHCLmX0u\naSsyeM6OzBJBrwTOBY6uZhsHmdnyHMpcZwz4Ni8jDfVHEJxzroHKpTfmV8CHwKHx2bfuZvZ8Bust\nNrMp8ffVQGUi6MRllprZRGB9tuUqCDO+pg1H8HShS+Kcc64GufTGPA94BOgeXw9L+nmW2+hH+kTQ\nyQx4XtJkSSOr2W6dJ4JOthlfsycT6cXH9bI/55xz2cvlnt0ZwF5mdrWZXQ3sDfwk05UzSQRdjaFm\nNhg4HDhb0v7JC9RLIugUWmAczMsUFdXbLp1zzmUhl2AnqmZALidNZpQNK2aQCLo6ZrYo/lwKPEkY\n467w7ruPdRRzOM9RUVHowjjnnEsll2B3P/A/SddKupYwyOp96VbKJBF0Deu2jZ1akNQWOAR4N9uC\n14nTT6cl6zmcZ2nN2kKXxjnnXApZ98Y0s1slvQoMjZNGmFkmw5imTQQtqQcwCegAVEg6HxgAdAOe\njMmhWwJ/M7Pnsi17XWkBdGQVB/IqPXsezqLCDD7knHOuGhkHO0ldEv6cF18b5pnZyprWzzAR9BLC\nWHXJVgG7ZVrWerfNNpTNnc9RjGHsJ4cXujTOOeeSZFOzm0zoEVkZsCqfKlP8fZs8lqtx+fBDpBYc\nw5OczR/I7fFF55xzdSXjb2Uz61+XBWnsijB68ClDGc/xx+/Po48WukTOOecq5dJBxaXSujUViOE8\nxWOPFbowzjnnEtVbsMtDbszDJM2WNEfSZfVV7ox9/TUVtOD7PAH4MwjOOdeQ1GfNrjI35gDCg+hn\nSxqQtExlbsxbEidKKgLuIjx17UYHAAAgAElEQVRQPgA4IcW6BdeScrZlLrszlQceKHRpnHPOVcol\nXdjvJA3Mdr1a5sYcAswxs7lmtg4YTQMdHd2AI/kPI0YUuiTOOecq5VKzmwncI+l/ks6S1DHbDeSQ\nG7MnsCDh74UkBcq43XrPjVmFGetpGZsynXPONRS5jHpwr5ntC5wK9AOmS/qbpIMyWb8WuTEzKVtB\ncmMmakUZJUxje2YzdWr65Z1zztW9nO7ZxXtoO8XXcmAacIGk0WnWyzU35iKgd8LfveK0ButInmb3\n3QtdCuecc5DbPbvfA7OB7wK/MbM9zOwmMxtOaJqsbr2cc2MCE4HtJfWX1Ar4ETAm27LXCzO+oZhj\neLLQJXHOORflkupjOnClmX2ZYl5NIxHknBvTzFbFgWLHAkXAKDN7L4ey14vWrGdf3qAnC6haIXXO\nOVcIuQS7B4BjJA0ldD583cyeBDCzL6pbqZa5MTGzZ4BncihvQbTAOJznkH6CWfrlnXPO1Z1c7tnd\nBZwFvEMYZuenku7Ka6kau7ffZp03ZTrnXIORS83uYGBns1BfkfRXoME2KRZESQmtWM8wXqQzK4Cu\nhS6Rc841a7nU7OYQ77NFveM0l6QV6zmUsSijcdydc87VlVyCXXtgpqRX4yCuM4AOksZIapg9JAvh\n/vtZR0uO5t+FLolzzjV7six7T0g6oKb5ZjaumvV6Aw8CWxI6ttxjZrcnLSPgdsJjDV8BP65MMSap\nnHCfEOBjM/teTeUoLS21SZMmpX9DdcgkvmJzuvMpX1q7gpbFOecyIWmymZUWuhz5lvU9OzMbJ2lL\nYM84aYKZLc1g1cpE0FMktQcmS3rBzGYkLHM4sH187QX8Kf4EWGtmJdmWt5AEtOUrDuIVevcezoIF\naVdxzjlXB3J5qPx4YAJwHHA88D9Jx6ZbL5NE0ITkzg9a8BbQSdJW2Zaxwdh7b8oo4ns8xcKFhS6M\nc841X7ncs/s/YE8zO83MTiU8SH5VNhuoIRF0TQmf28Qkz29JOrqa7RY2EXSyN99EGN9jDKKs0KVx\nzrlmK5dg1yKp2XJFNtupRSLovrEd+UTgNknbJi/QEBJBJyuigh58yrd4i3POKXRpnHOuecol2D0n\naaykH0v6MfA0GWY2ySARdLUJn82s8udc4FVqyMPZoBQXU4E4kv9wlz9675xzBZHLED8XA38Gdo2v\ne8zs0nTrZZgIegxwqoK9gS/MbLGkzpJax+10I+TZnFHNNhqWdesop4ij+VehS+Kcc81WVr0x49A+\nL5rZQZD1CKVpE0ETaojfJTyk/hVQOd73zsCfJVUQAvSNSb04G7RiytiJ2QzgXaZNG8RuuxW6RM45\n17xkFezMrFxShaSONSV9rmbdTBJBG3B2iunjgV2y2V9D9F2epaRkkCeGds65epZLbsw1hNrZC8CG\nYX7M7Ny8laqpMWOdWnE0/+IWLi50aZxzrtnJJdg9waZNmF5XSaMV69mHN32MO+ecK4Bcgl2nFGm+\nzstTeZq0FhiH8jzSGd6U6Zxz9SiXRw9OSzHtx7UsR9M3dSrrKPZemc45VwDZPAx+gqSngP6VIxzE\n1yvAygzW7y3pFUkzJL2XqjYYHzm4Q9IcSdMlDU6Yd5qkD+IrVcBt2HbbjVas5zu8QHuy6tvjnHOu\nlrJpxhwPLAa6Ab9LmL4amJ7B+jkngpbUBbgGKCXcH5wsaYyZfZZF+RuENnzDMF6kuPgHrF9f6NI4\n51zzkHHNzszmm9mrZraPmY1LeE0xs7SJH2uZCPpQ4AUzWxkD3AvAYZmWvcE4+2zW05KjGEOZp8p0\nzrl6k8uoB9+PTYlfSFolabWkbHJc5pIIuqYE0YnbbViJoJPdeSdFlDGcp2jJukKXxjnnmo1cOqjc\nDHzPzDqaWQcza29mHTJduRaJoNNqiImgk7UAuvAZ+/G6Z1Jxzrl6kkuw+9TMZuays1okgq42QXSj\n07kz5bRgOE8xPZM7nc4552otl2A3SdI/Yu/M71e+0q1Um0TQwFjgkJgQujNwSJzW+KxcCRjf5wmg\notClcc65ZiGXh8o7EJI0H5IwzUifGDrnRNBmtlLSL4GJcb3rzSzt4w4NVRFGXz5mV95B2s0fMHfO\nuToma6LftKWlpTZp0qRCFyM1CQPu4SecxT0e7JxzDYakyXGg7CYlm4fKH034/aakec/ns1BNnhkC\nRvAAXViOahwLwjnnXG1lc89u+4Tfv5M0r2F2fWzgWrGen3JPoYvhnHNNXjbBrqbGNm+Iy1Zsu7yI\nWyjmG7p2LXB5nHOuCcsm2G0uaXdJewCbxd8HV/6dbmVJoyQtlfRuNfM7S3oy5sScIGlQwrx5kt6R\nNFVSA70Rl5sufMbxPBY6aTrnnKsT2QS7xcCtwC3Akvj77xL+TucBak7xdQUw1cx2BU4Fbk+af5CZ\nlTSpG6fTpgFwBb8GKrjhhsIWxznnmqqMHz0ws4NqsyMzey2mCavOAODGuOwsSf0kbWlmn9Zmvw3a\nrrsCMIBZ7M9/ueKKA7j88gKXyTnnmqBcHiqvK9OA7wNIGgL0JWRKgXBP8HlJkyWNLFD56sYOO2DA\nZXi1zjnn6kpDCnY3EkY5mAr8HHgbKI/zhprZYMIQQGdL2j/VBhp8IuhUZs9GwGGMZVvm+GMIzjlX\nBxpMsDOzVWY2wsxKCPfstgDmxnmL4s+lwJPAkGq20eATQdfkwirDBDrnnMuXXIb42VdS2/j7yZJu\nldS3tgWR1ElSq/jnmcBrZrZKUts42Ctxv4cAKXt0NlrxIfPTGUUnVnrtzjnn8iyXmt2fgK8k7QZc\nCHwIPJhuJUl/B94EdpS0UNIZks6SdFZcZGfgXUmzCc2V58XpWwKvS5oGTACeNrPncih3g9eadYzk\nL4UuhnPONTlZ58aUNMXMBku6GlhkZvdVTqubIuamQefGrI7EZ3SiO0spo9hzZjrn6l2zz42ZYLWk\ny4GTgacltQCK81us5qszn3MaDwCw336FLYtzzjUVuQS7HwLfAGeY2RLC4wG/zWupmquHHgLgD5xL\nB77g9dcLXB7nnGsicqrZAbeb2X8l7QCUAH/Pb7GaqZNPBmAzvuZ+fgzgnVWccy4Pcgl2rwGtJfUE\nnicMyPpAPgvVrMUbdcfwL/bjNQD+7pcSzjlXK7kEO5nZV4RsJ380s+OAQWnWqW0i6MMkzZY0R9Jl\nOZS5cenaFQFPcgzFrOPEEwtdIOeca9xyCnaS9gFOAp7OYjsPkEMiaElFwF2ExxEGACdIGpBDuRuP\n5csB6MpKbuISwJsznXOuNnIJducDlwNPmtl7krYBXkm3kpm9BtQ0kM0A4OW47Cygn6QtCdlS5pjZ\nXDNbB4wGjsqh3I1LbM48jzvYkZkFLoxzzjVuWQc7MxtnZt8D7pLULgahc/NQluoSQfcEFiQstzBO\n20SjzI2ZRguM/zAcUeG1O+ecy1Eu6cJ2kfQ28B4wI45EMDAPZakpEXRGGntuzE3E2t12fMjPuQPw\n5kznnMtFLs2YfwYuMLO+ZtaHkDKs1jmuakgEvQjonbBorziteZg+HYBbuZAefAJAu3aFLJBzzjU+\nuQS7tma24R6dmb0KtK1tQapLBA1MBLaX1D/O/xEwprb7azR22QWAIiqYRCktKOPLL+GCCwpcLuec\na0RyCXZzJV0VRxLvJ+lK4lA8Nck1EbSZlQHnAGOBmcCjZvZeDuVuvGJzZk8W8zLfBuD3v4d3m9bY\nD845V2dySQTdGbgOGEoYQfy/wHVm9ln+i5e7RpkIOp14w+5WfsGF3ArgyaKdc3nliaDZ8Mzb/5nZ\nuWY22Mz2MLPzG1qga7JuvhmAX/B7TiLk0fQOK845l15Wwc7Mygk1OlcIF18MnTsj4CFOpZQJgAc8\n55xLp2UO67wtaQzwGPBl5UQzeyJvpXLVW7kSJAS8ybfow3wW0xPJmzSdc646uQS7NsAK4OCEaQZ4\nsKsvZiDRknLeZ0e6s4S1tPOA55xz1cg62JnZiFx2JGkUcCSw1Mw2SRwtqSPwMNAnlusWM7s/zisH\n3omLfhwzuDRvMeC140sW05PuLGEdm3nAc865FDK+Zyfpt5J+mmL6TyXdmMEmHqDmRNBnAzPMbDfg\nQOB3Cc/drTWzkvjyQFcpRrWOrOJzOtMpph71e3jOOVdVNh1UDgbuSTH9L4QaW40ySARtQHtJAtrF\nZcuyKF/zVFEBwGZ8wzK604ePAA94zjmXKJtg19pSPJRnZhVAPr5a7yQ8WP4JocnyvLhtgDYxwfNb\nko6ubgNNMRF0Wgntli0p5yO2ZQ8mbpj12GOFLJxzzjUM2QS7tZK2T54Yp63NQ1kOBaYCWwMlwJ2S\nOsR5feNDjicCt0naNtUGmlwi6GzEgNcCYyJDOJJ/A3D88dAilzw5zjnXhGTzNXg18KykH8eRD3aR\nNIIwgOvVeSjLCOAJC+YAHwE7AZjZovhzLvAqsHse9tf0xIAnYAxHcwXXb5jszZrOueYs42BnZs8C\nRwMHETqbPEDoSPIDM3smD2X5GELixzho646EPJydJbWO07sB+wIz8rC/pikh4P2aa5jOIFrEW58S\nXHJJAcvmnHMFknVuzJx3FBJBHwh0Az4FrgGKAczsbklbEwLoVoTv6hvN7GFJ3yIMK1RBCM63mdl9\n6fbXJHNjZqNTJ/jiCwDWU8Qg3uX9UFEG/PEE51xqTTU3Zi4PlefEzE5IM/8T4JAU08cDu9RVuZqs\nzz+HGTNg4ECKKWc2O3MDF3MFIb+mFGbvvHOBy+mcc/XAuy40ZQMGVKnCXc5vWUhPWsf+RAMG+L08\n51zzkJdgl/Dwt2uIzGDvvQHoySesZXN+yCMbZkuw776FKpxzztW9rIOdpFcl9Uv4ewjEB7tcw/Xm\nm1U6r4zmZJbThW4sBWD8eK/lOeearlxqdjcAz0n6maRfA3cTHhuokaRRkpZKSjm+tqSOkp6SNE3S\ne/Gxhsp5p0n6IL5Oy6HMrpIZtAoV8a58xjK25DF+QOj/EwKeBz3nXFOTU29MSQcCLwDLgd3NbEkG\n6+wPrAEerCYR9BVARzO7VNIWwGygByF12CSglJBSbDKwR7oBY5t9b8xMJES1clownH/zbELmt5Yt\nYf36QhTMOVcoTbU3Zi7NmFcBfwD2B64FXpV0RLr1apEb81DgBTNbGQPcC9ScUNplygxuuw2AIip4\nhuEspwt9mQdAWVmIh/vtV8AyOudcHuTSjNkVGGJmb5rZnwnB6Pw8lKW63Jg9gQUJyy2M01w+nHde\nCHoxp1hXPmMe/ZnBjrQjPKf3+utVUnA651yjk3WwM7PzzWxtwt/zzew7eShLTbkxM9IsE0HnS3l5\nlWi2M++zmk48xREUEdoyW7SA4uJCFdA553KXSzPmFpJukfSMpJcrX3koS3W5MRcBvROW6xWnbaJZ\nJ4LOF7MqQe9InmE9rbiAmwHb0LTpnHONSS7NmI8AM4H+wHXAPPLz6EHK3JjAWOCQmCOzMyHLytg8\n7M/VxAxmzgTCowq/41IWs9WGRxWkDZ06nXOuwcvpnl3MTbnezMaZ2emEgV1rFHNjvgnsKGmhpDMk\nnSXprLjIL4FvSXoHeAm41MyWm9nKOG9ifF0fp7m6ttNOIegdFvoD9eBTlrIllxAGpl+/3mt5zrnG\nIetHDyS9ZWZ7SxoL3EHoUPJPM0s5xlyh+KMHdSAhsi1mS3ZhOivoDkBRUei96Zxr3PzRg41+Jakj\ncCFwEXAv8Iu8lso1TGbwox8BsBWfsowtuYwbgNC/RYL33y9kAZ1zLrV6G+KnvnnNro4l1PKW0o3d\nmMYStt4wrYmeVs41eU21ZpfxED+S7qhpvpmdW/viuEbDDE44AUaPpjvL+YSe3MxFXMZNQAskuPpq\nuO66QhfUOeeyqNlJWge8CzxKuE9XpWuCmf0176WrBa/Z1aOEWt4XdKCEKcxj4y3cYcPghRcKUTDn\nXLaaas0um3t2WwH3EB7+PoUwyvi/zeyvmQS6DBJBXyxpany9K6lcUpc4b56kd+I8j2ANjRlcdRUA\nHVnFXLbjHxyPKAfgxRdDPNxqq0IW0jnXnOWaCLoX8CPgAsIjAg9lsE6NiaCTlh0O/MLMDo5/zwNK\nzWx5pmX0ml2BJNTyDHiKIziWx1lP6yqL+T095xomr9lFkgYD5wEnA88SRiFIK4NE0IlOAP6ebdlc\nA2AGTz0FhHbu7/E062jDW+xFW1ZtWKxyKKHnnitQOZ1zzUrGwU7S9ZImE2pz4wg1rTPMbEY+CyRp\nc8KoBo8nTDbgeUmTJY2sYV3PjdkQHHlkCHoJzyHsxQTW0JE5bLNhVAWAww8PQc9zbjrn6lI2Nbsr\ngU7AboQBXKdImh7vpU3PY5mGA28kZUkZamaDgcOBs2OT6CY8N2YDs/32mwS9bfmIefRnHS25gN9S\nOWhsZc5NCY5IO2CUc85lJ5tg15+QFuzI+BoeX5W/58uPSGrCNLNF8edS4ElgSB735+paZdBLuFFX\nTDm/4xKMIqZQwhZsHP/3mWc2Br777itEgZ1zTU3GwS4O5VPtKx+FiZlZDgD+nTCtraT2lb8TEkGn\n7NHpGoHKoNe374ZJuzONpWxFGS24hfMRG/OOnXnmxsB3882FKLBzrinIJV1YTjJIBA1wDPC8mX2Z\nMG1L4HVJ04AJwNNm5t0aGrt580LQ++CDDZOKMC7kdiooZi2t+Rl3UtnMCXDppRsD3z771H+RnXON\nl6cLcw3HkCEwcdPRolbTlgu4lXs5k+quz5roaexcvfNHDyJJ52UyzbmsTZiwsZmzbdsNk9vzJX/h\npxhFrKMlN3EJRayrsmpljU+CQw6p74I75xq6XJoxT0sx7ce1LIdzVa1ZszHwtdyYwrWYci7ht5TR\nmgrEKxxAXz6ssuoLL1QNfud61lbnmr1snrM7QdJTQH9JYxJer5D5w+LOZW/9+o2Br0+fDZMFHMhr\nzGM7DLGGzbmEG6p0cAH4wx+qBr/TT6/n8jvnCi6bmt0U4HfArPiz8nUhIV+mc3Vv/vyNge/yy6vM\nastabuIKKijGEIvYilO4n8ROLgD33181+G29Nc65Ji6bYPd3M3sV+NDMxiW8pphZ2jGqa5kI+jBJ\nsyXNkXRZFmV2TdlvfrMx8JlBmzZVZm/NEh7kdIwiDDGPPvyYUZvU/BYvrhr8JPjjH+vzjTjn6lo2\nwa6VpBOBfSR9P/mVwfoPENKApWRmvzWzEjMrAS4HxpnZSklFwF2E7CkDgBMkDcii3K65WLt2Y+D7\n6KNNZvdlAfdzxoaa32racjMX0Y4vNln27LM3DYDjx9fHm3DO1YVsgt1ZwH6ElGHDk15Hplu5Fomg\nhwBzzGyuma0DRgNHZVFu1xz161e11pci+LXjKy7md6ymE4aoQExgD77HkyQ3fQLsu++mAfDee+v+\nrTjnai/r5+wknWFmOSVxktQP+E9NQ/zERNALge1ize5Y4DAzOzPOPwXYy8zOSbHuSGAkQJ8+ffaY\nPz8viV1cU9W2LXz1VY2LrKeIsRzGTVzM6+xHJteH/fqljK3ONQr+nN1GD0k6V9I/4+vnkvKZsz5V\nIuiMeCJol5Uvv6xa+/vlLzdZpJhyjuRp/suBG+79hQA4jOH8e5P7fxCSwyTXACU4ytsjnCuYXILd\nH4E94s8/AoOBP+WxTMmJoBcBvRP+7hWnOZdfV15ZNfiZwf6bDrDRkgoO4SXGcPSG+38ViLn04yqu\no0c1p+eYMamD4G671fUbc87lEuz2NLPTzOzl+BoB7JmPwqRKBA1MBLaX1F9SK0IwHJOP/TmX1rhx\nmwbAY47ZZDEB/ZnP9VzLYnphCEOU0YK3GMJZ/IkOfJZyF9Onpw6CLVvC330IY+fyIpdgVy5p28o/\nJG0DlKdbKddE0PGxhnOAscBM4FEzey+HcjuXH088sWkAXLAAWmz6cSrC2IuJ/Imf8QVdqgTBt9mV\ny/kVfUh9g6+8HE48MXUgbN0aHnusrt+oc01HLh1Uvg3cD8wlXND2BUaY2Sv5L17uPBG0axCuugp+\n9ausVllKN17mYB7jWMZyKF/SIevdfvvb8OKLWa/mXJPtoJLTqAeSWgM7xj9nm9k3eS1VHniwcw3a\ngw/CiBFQsekjDtUxYAG9eJUDeIYjeZ5hfEa3rHddVATHHgujR2e9qmsGmn2wk7QnsMDMlsS/TwV+\nAMwHrs2l92Rd8mDnGq0BA2DmzKxXW0knplHCyxzI8xzCZPagnFY5FaFXr1Az3HHH9Mu6psWDnTQF\nGBaffduf8HD3z4ESYGczO7buipk9D3auSdpxR3j//axXqwCW0IOp7MZ/Gco4DmQSe7Ke1jkVo0UL\n2GYbeOSRMAyhazo82EnTzGy3+PtdwDIzuzb+PTWm+WowPNi5Zueaa0K+0LK0qWo3YcAqOjCbHZjI\nnrzOvrzJvsynD7n1Yws23xy+8x144AHo1Cnnzbh65MEuJHAuMbMySbOAkTEFGJLerSkrSlxmFCGt\n2NLqlpV0IHAbUAwsN7MD4vR5wGpCr8+yTP4RHuycS/KLX8Bdd4Uhk3JQAaygKx+yLW9TwkSGMIEh\nzGInyqldXok2bWDgQLjtNhg6tFabcrXkwU76P+C7wHKgDzDYzEzSdsBfzWzfNOvvD6wBHkwV7CR1\nAsYTUoN9LKm7mS2N8+YBpWa2PNM35sHOuSzNmhWqYQsX1moza2nNJ/RkNjswld2ZzO5MpYSP6I/R\nMv0GaiCFLG877wyXXgo/+EGtNudSaPbBDkDS3sBWJDwLJ2kHoJ2ZTclg/X5UkxtT0s+Arc3syhTz\n5uHBzrnCe/NNOOWUkPwzi56kqXxDMUvpzkf0YxY78w67MI1deY9BrKQr4cmm2ikuhq5dYdAguPBC\nOKzacVdcJQ92+dhZzcGusvlyINAeuN3MHozzPgI+I9xa+LOZ3VPN9j0RtHMNwW23hfuHy5eHh+5r\nwYCvac0yuvMxvZnD9sxgR2YykJnszDz61boZtVLLltCuXRjQd8894Sc/CaNdNCce7PKxs5qD3Z1A\nKfBtYDNCtpUjzOx9ST3NbJGk7sALwM8r7xdWx2t2zjUC77wDZ50FU6emHYEiUwasoR1L2YKF9OQj\n+vMBOzCbHfiAHZhLf9bQMS/7kkLtsX176NYt9FDde+/QvDpwYF52Ue882OVjZzUHu8uAzczsmvj3\nfcBzZvZY0nLXAmvM7Jaa9uXBzrkmZuFCuPhiePXVUGPModdpddZTxCo6sJxuLGJrPqYvc+nPXLZj\nDtswj358Sg9q0zO1UmW6t1atQi2yc2fYaivo3x/22w/22ScME9Wydrc3a1E+D3a131nNwW5n4E7g\nUKAVMIGQ9PkjoIWZrZbUllCzu97MnqtpXx7snGvm5syBG2+E//4XPvkk1BxreZ8xkQFrabMhSC6h\nBwvozXz6Mo9+fEwfPqY3i+jJ17St1b6KikJwbNUq9Fxt2xY6dIA+fcL9yO7doUePEDR79oTevcOy\nufBgV9sdhUTQBwLdgE+Bawj36DCzu+MyFwMjCL2c7zWz22Ki6SfjZloCfzOzX6fbnwc751zWPv4Y\n/vIXeOml0Anniy9g3bqQlTuPymnBWjZjFR34TF1YSnc+YWsWF/VioXqxwHqx0Lbmk4qt+dS6U6ZW\nWd36bN0avv46t7J5sGtkPNg55+rNsmXwz3+GJtZZs2Dp0jA4cFlZeK6xvLzWHXWADcNelBcVs7ZF\nW77cvBtfde7N50VdWKmurKArS60733TbmgvH55bUyoNdI+PBzjnXKHz1FbzxRnisY/nycG9yyRJY\nsQJWrw7zW7QIAXPduhBAKypqbpJt0wbWrs2pOE012BXoFqhzzjlgY06173wn922sWgWLFoV7k0uW\n5K9sTYgHO+eca+w6dAivnXcudEkarNr3o82QpFGSlsYcm9Utc6CkqZLekzQuYfphkmZLmhMfUXDO\nOecyVm/BDngAqDZZT8yN+Ufge2Y2EDguTi8C7gIOBwYAJ0gaUOeldc4512TUW7CLGU9qGuD1ROAJ\nM/s4Lr80Th8CzDGzuWa2jjCO3lF1WljnnHNNSn3W7NLZAegs6VVJk+NI6AA9gQUJyy2M0zYhaaSk\nSZImLVu2rI6L65xzrrFoSB1UWgJ7kJAbU9Jb2WwgJoi+B8KjB3kvoXPOuUapIQW7hcCKOHTQl5Je\nA3aL03snLNcLWFSA8jnnnGukGlIz5r+BoZJaStoc2AuYCUwEtpfUX1IrQr7MMQUsp3POuUam3mp2\nibkxJS0kKTemmc2U9BwwnY25Md+N654DjAWKgFFm9l59lds551zj5+nCnHPObdBU04U1pGZM55xz\nrk54sHPOOdfkebBzzjnX5Hmwc8451+Q1mETQMQn0FzER9FRJVyfMmyfpnTjde50455zLSn0+VP4A\ncCfwYA3L/NfMjqxm3kFmtjzvpXLOOdfkNaRE0M4551ydaGj37PaRNE3Ss5IGJkw34PmYIHpkdSt7\nImjnnHOpNKTcmFOAvma2RtJ3gX8B28d5Q81skaTuwAuSZsWaYhWeCNo551wqDaZmZ2arzGxN/P0Z\noFhSt/j3ovhzKfAkYYw755xzLiMNJthJ6iFJ8fchhLKtkNRWUvs4vS1wCJCyR6dzzjmXSoNJBA0c\nC/w/SWXAWuBHZmaStgSejHGwJfA3M3uuvsrtnHOu8au3YGdmJ6SZfyfh0YTk6XMJ49o555xzOWkw\nzZjOOedcXfFg55xzrsnzYOecc67J82DnnHOuyWssiaAPkzRb0hxJl9VXmZ1zzjUN9VmzewA4LM0y\n/zWzkvi6HkBSEXAXcDgwADhB0oA6LalzzrkmpTEkgh4CzDGzuWa2DhgNHJXXwjnnnGvSGlJuTIiJ\noIFPgIvM7D2gJ7AgYZmFwF6pVo5JoisTRa+RNLsWZekGNMchhfx9Ny/+vpuXTN533/ooSH1rSMGu\npkTQGUlMBF1bkiaZWWk+ttWY+PtuXvx9Ny/N9X1DA+qNWUMi6EVA74RFe8VpzjnnXEYaTLCrLhE0\nMBHYXlJ/Sa2AHwFjCn9bBgoAAAUESURBVFdS55xzjU2DTwQNlEk6BxgLFAGj4r28upaX5tBGyN93\n8+Lvu3lpru8bhXjinHPONV0NphnTOeecqyse7JxzzjV5HuySNKfUZKlSuEnqIukFSR/En50LWcZ8\nk9Rb0iuSZkh6T9J5cXqTft8AktpImiBpWnzv18Xp/SX9L57z/4gdwZoUSUWS3pb0n/h3k3/PAJLm\nSXonpmCcFKc1+XM9FQ92CZpharIH2DSF22XAS2a2PfBS/LspKQMuNPv/7d1NiFV1GMfx76+SMCcS\nxSSUEnPRC9hIIJYGNlEESbnoBVKRaOkiF1EYtRFctElbBAm1mMggM6e2mYnkojdNKtJNUaCYs9Fq\ngqT01+L8b96ZJqKr1zv87+8Dwz3n3MPh/zD/O895mfs8vgVYBmwov+Pa4wY4AwzZvg0YBO6XtAx4\nEdhqexFwCniyh2PslqeAI23r/RBzy92lBGPr+3X9MNf/IcluvL4qTfYvJdweAobL8jCw+pIOqsts\nn7B9qCz/SvMHcB6Vxw3gxlhZnVZ+DAwBu8r26mKXNB94AHitrIvKY/4P1c/1ySTZjTdZabJ5PRpL\nr8y1faIs/wTM7eVguknSAmAJ8Cl9Ene5nXcYGAX2AN8Bp23/WXapcc5vA54BzpX12dQfc4uBDyQd\nLOUUoU/m+kRTqVxYTDG2LanK76ZIGgDeBTba/qXUMwDqjtv2WWBQ0kxgBLipx0PqKkmrgFHbByWt\n7PV4emCF7eOSrgX2SDra/mbNc32iXNmNl9JkcFLSdQDldbTH47noJE2jSXQ7bO8um6uPu53t08A+\n4A5gpqTWiW9tc3458KCkH2geSwwBL1N3zH+zfby8jtKc3Cylz+Z6S5LdeClN1sS7viyvB97v4Vgu\nuvK85nXgiO2X2t6qOm4ASXPKFR2SpgP30jyz3EdTwQgqi932JtvzbS+g+Tx/ZHsNFcfcImmGpKtb\ny8B9wDf0wVyfTCqoTFA6LmzjfGmyLT0eUte0l3ADTtKUcHsP2AlcD/wIPGq7kz6EU5KkFcDHwNec\nf4bzHM1zu2rjBpC0mOYfEi6nOdHdaXuzpIU0Vz2zgC+BtbbP9G6k3VFuYz5te1U/xFxiHCmrVwBv\n2d4iaTaVz/XJJNlFRET1chszIiKql2QXERHVS7KLiIjqJdlFRET1kuwiIqJ6SXYRU5Ckla0K/RFx\n4ZLsIiKiekl2ERdA0trSI+6wpO2l0PKYpK2lZ9xeSXPKvoOSPpH0laSRVh8xSYskfVj6zB2SdGM5\n/ICkXZKOStqh9gKeEfG/JNlFdEjSzcBjwHLbg8BZYA0wA/jC9q3AfprKNABvAM/aXkxTwaW1fQfw\nSukzdyfQqki/BNhI01txIU2dx4joQLoeRHTuHuB24PNy0TWdpqjuOeDtss+bwG5J1wAzbe8v24eB\nd0rtwnm2RwBs/w5QjveZ7WNl/TCwADjQ/bAi6pNkF9E5AcO2N43bKL0wYb9Oa/K112o8Sz6vER3L\nbcyIzu0FHi69wpA0S9INNJ+rVkX9x4EDtn8GTkm6q2xfB+wv3dKPSVpdjnGlpKsuaRQRfSBnihEd\nsv2tpOdpOkFfBvwBbAB+A5aW90ZpnutB007l1ZLMvgeeKNvXAdslbS7HeOQShhHRF9L1IOIikzRm\ne6DX44iI83IbMyIiqpcru4iIqF6u7CIionpJdhERUb0ku4iIqF6SXUREVC/JLiIiqvcXyVYLyZou\nf2QAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L58VewLbf0yL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "73fb099e-aebe-444c-f070-c7e48b1f0745"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAEMCAYAAAAs8rYIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nzjkXIU/+zjkXIU/+zjkXIU/+zjkXof8Df69PayBoPi0AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "i0XG75fugqvU",
        "colab_type": "text"
      },
      "source": [
        "**Model 3: Input - Sigmoid(BatchNormalization(512)) - Sigmoid(BatchNormalization(128))- Sigmoid(output)** "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XT5I3hAOgGdz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# https://www.tensorflow.org/api_docs/python/tf/nn/batch_normalization\n",
        "# https://r2rt.com/implementing-batch-normalization-in-tensorflow.html\n",
        "epsilon = 1e-3\n",
        "def multilayer_perceptron_batch(x, weights, biases):\n",
        "    # Use tf.matmul instead of \"*\" because tf.matmul can change it's dimensions on the fly (broadcast)\n",
        "    print( 'x:', x.get_shape(), 'W[h1]:', weights['h1'].get_shape(), 'b[h1]:', biases['b1'].get_shape())        \n",
        "    \n",
        "    ############################################################\n",
        "    # Hidden layer with Sigmoid activation and batch normalization\n",
        "    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) #(x*weights['h1']) + biases['b1']\n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/moments\n",
        "    # Calculate the mean and variance of x.\n",
        "    batch_mean_1, batch_var_1 = tf.nn.moments(layer_1,[0])\n",
        "    \n",
        "    scale_1 = tf.Variable(tf.ones([n_hidden_1]))\n",
        "    beta_1 = tf.Variable(tf.zeros([n_hidden_1]))\n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/batch_normalization\n",
        "    layer_1 = tf.nn.batch_normalization(layer_1, batch_mean_1, batch_var_1, beta_1, scale_1, epsilon)\n",
        "    layer_1 = tf.nn.sigmoid(layer_1)\n",
        "    \n",
        "    print( 'layer_1:', layer_1.get_shape(), 'W[h2]:', weights['h2'].get_shape(), 'b[h2]:', biases['b2'].get_shape())        \n",
        "    \n",
        "    #####################################################################################\n",
        "    \n",
        "    # Hidden layer with Sigmoid activation and batch normalization\n",
        "    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) # (layer_1 * weights['h2']) + biases['b2'] \n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/moments\n",
        "    # Calculate the mean and variance of x.\n",
        "    batch_mean_2, batch_var_2 = tf.nn.moments(layer_2, [0])\n",
        "    \n",
        "    scale_2 = tf.Variable(tf.ones([n_hidden_2]))\n",
        "    beta_2 = tf.Variable(tf.zeros([n_hidden_2]))\n",
        "    \n",
        "    layer_2 = tf.nn.batch_normalization(layer_2, batch_mean_2, batch_var_2, beta_2, scale_2, epsilon)\n",
        "    layer_2 = tf.nn.sigmoid(layer_2)\n",
        "    print( 'layer_2:', layer_2.get_shape(), 'W[out]:', weights['out'].get_shape(), 'b3:', biases['out'].get_shape())        \n",
        "    \n",
        "    ######################################################################################\n",
        "    \n",
        "    # output layer with Sigmoid activation \n",
        "    out_layer = tf.matmul(layer_2, weights['out']) + biases['out'] # (layer_2 * weights['out']) + biases['out']    \n",
        "    out_layer = tf.nn.sigmoid(out_layer)\n",
        "    print('out_layer:',out_layer.get_shape())\n",
        "\n",
        "    return out_layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6PMRhOWlhc_J",
        "colab_type": "text"
      },
      "source": [
        "**Model 3 with Adam Optimizer**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "otkyZim8hU-H",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1380
        },
        "outputId": "a74662a0-c1e7-4ee7-829d-efde82598e99"
      },
      "source": [
        "# Since we are using sigmoid activations in hiden layers we will be using weights that are initalized as weights_sgd\n",
        "ybatch = multilayer_perceptron_batch(x, weights_sgd, biases)\n",
        "\n",
        "# https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits\n",
        "cost_batch = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = ybatch, labels = y_))\n",
        "# https://github.com/amitmac/Question-Answering/issues/2\n",
        "# there are many optimizers available: https://www.tensorflow.org/versions/r1.2/api_guides/python/train#Optimizers \n",
        "optimizer_batch_adam = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost_batch)\n",
        "optimizer_batch_sgdc = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost_batch)\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use AdamOptimizer\n",
        "            _, c, w = sess.run([optimizer_batch_adam, cost_batch, weights_sgd], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_batch, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(ybatch,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x: (?, 784) W[h1]: (784, 512) b[h1]: (512,)\n",
            "layer_1: (?, 512) W[h2]: (512, 128) b[h2]: (128,)\n",
            "layer_2: (?, 128) W[out]: (128, 10) b3: (10,)\n",
            "out_layer: (?, 10)\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "Epoch: 0001 train cost=1.697667968 test cost=1.689333970\n",
            "Epoch: 0002 train cost=1.554137292 test cost=1.550786987\n",
            "Epoch: 0003 train cost=1.516122974 test cost=1.515506031\n",
            "Epoch: 0004 train cost=1.500450389 test cost=1.501982761\n",
            "Epoch: 0005 train cost=1.491390617 test cost=1.494789489\n",
            "Epoch: 0006 train cost=1.486090121 test cost=1.491460966\n",
            "Epoch: 0007 train cost=1.482240193 test cost=1.488212924\n",
            "Epoch: 0008 train cost=1.478775082 test cost=1.486447946\n",
            "Epoch: 0009 train cost=1.477364177 test cost=1.485353771\n",
            "Epoch: 0010 train cost=1.475260521 test cost=1.484205269\n",
            "Epoch: 0011 train cost=1.473408004 test cost=1.483065278\n",
            "Epoch: 0012 train cost=1.472226460 test cost=1.482310045\n",
            "Epoch: 0013 train cost=1.471575944 test cost=1.482188640\n",
            "Epoch: 0014 train cost=1.470511063 test cost=1.481614506\n",
            "Epoch: 0015 train cost=1.469464036 test cost=1.481465636\n",
            "Epoch: 0016 train cost=1.469192652 test cost=1.481757392\n",
            "Epoch: 0017 train cost=1.469047668 test cost=1.480887899\n",
            "Epoch: 0018 train cost=1.467809549 test cost=1.480477857\n",
            "Epoch: 0019 train cost=1.467257042 test cost=1.480175611\n",
            "Epoch: 0020 train cost=1.467121260 test cost=1.480760045\n",
            "Epoch: 0021 train cost=1.467227978 test cost=1.480665346\n",
            "Epoch: 0022 train cost=1.466466006 test cost=1.480212677\n",
            "Epoch: 0023 train cost=1.466440056 test cost=1.480405730\n",
            "Epoch: 0024 train cost=1.465850416 test cost=1.479988253\n",
            "Epoch: 0025 train cost=1.465778477 test cost=1.479545294\n",
            "Epoch: 0026 train cost=1.465343836 test cost=1.480624137\n",
            "Epoch: 0027 train cost=1.465845668 test cost=1.480385397\n",
            "Epoch: 0028 train cost=1.465137188 test cost=1.479238033\n",
            "Epoch: 0029 train cost=1.464551811 test cost=1.479325228\n",
            "Epoch: 0030 train cost=1.465120443 test cost=1.479478983\n",
            "Epoch: 0031 train cost=1.464934185 test cost=1.479849409\n",
            "Epoch: 0032 train cost=1.464759494 test cost=1.478756195\n",
            "Epoch: 0033 train cost=1.464443506 test cost=1.479164549\n",
            "Epoch: 0034 train cost=1.464591265 test cost=1.479352745\n",
            "Epoch: 0035 train cost=1.464246520 test cost=1.478956791\n",
            "Epoch: 0036 train cost=1.463762583 test cost=1.478841414\n",
            "Epoch: 0037 train cost=1.464256600 test cost=1.479189302\n",
            "Epoch: 0038 train cost=1.464199858 test cost=1.479491770\n",
            "Epoch: 0039 train cost=1.463315068 test cost=1.478288112\n",
            "Epoch: 0040 train cost=1.463873339 test cost=1.479257436\n",
            "Epoch: 0041 train cost=1.463856849 test cost=1.479242996\n",
            "Epoch: 0042 train cost=1.463845836 test cost=1.478036420\n",
            "Epoch: 0043 train cost=1.463490822 test cost=1.479017681\n",
            "Epoch: 0044 train cost=1.463730200 test cost=1.479598890\n",
            "Epoch: 0045 train cost=1.463697612 test cost=1.478522304\n",
            "Epoch: 0046 train cost=1.463300974 test cost=1.478413649\n",
            "Epoch: 0047 train cost=1.463173054 test cost=1.478721926\n",
            "Epoch: 0048 train cost=1.463213968 test cost=1.479363461\n",
            "Epoch: 0049 train cost=1.463184417 test cost=1.479937710\n",
            "Epoch: 0050 train cost=1.463162822 test cost=1.478861100\n",
            "Accuracy: 0.9818\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdQAAAEWCAYAAADfB2bTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nXRcrKIohi_6",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "3ca0b890-44e7-4ca4-8a34-b7ac843099b6"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAEMCAYAAAAs8rYIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTR7JrWAAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o1Aklz2six1j",
        "colab_type": "text"
      },
      "source": [
        "Model 3 with Gradient Descent Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xoVDQUU-isF0",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1232
        },
        "outputId": "7e72d241-a251-405c-fd52-a197a2c15057"
      },
      "source": [
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use GradientDescentOptimizer\n",
        "            _, c, w = sess.run([optimizer_batch_sgdc, cost_batch, weights_sgd], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_batch, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax,np.arange(1.5, 2.4, step=0.05), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.5, 2.4, step=0.05), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(ybatch,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 0001 train cost=2.319629274 test cost=2.320529026\n",
            "Epoch: 0002 train cost=2.282610404 test cost=2.282330591\n",
            "Epoch: 0003 train cost=2.245107988 test cost=2.243535444\n",
            "Epoch: 0004 train cost=2.210171726 test cost=2.207523905\n",
            "Epoch: 0005 train cost=2.179199741 test cost=2.175606041\n",
            "Epoch: 0006 train cost=2.151590070 test cost=2.147420353\n",
            "Epoch: 0007 train cost=2.126844169 test cost=2.122033668\n",
            "Epoch: 0008 train cost=2.103964632 test cost=2.098933902\n",
            "Epoch: 0009 train cost=2.083487102 test cost=2.077934836\n",
            "Epoch: 0010 train cost=2.064841035 test cost=2.059234300\n",
            "Epoch: 0011 train cost=2.048718361 test cost=2.042765579\n",
            "Epoch: 0012 train cost=2.034457644 test cost=2.028455564\n",
            "Epoch: 0013 train cost=2.021926317 test cost=2.015886191\n",
            "Epoch: 0014 train cost=2.011060684 test cost=2.004742313\n",
            "Epoch: 0015 train cost=2.001073343 test cost=1.994707897\n",
            "Epoch: 0016 train cost=1.991944520 test cost=1.985514876\n",
            "Epoch: 0017 train cost=1.983510187 test cost=1.977063076\n",
            "Epoch: 0018 train cost=1.975801662 test cost=1.969175371\n",
            "Epoch: 0019 train cost=1.968412796 test cost=1.961801313\n",
            "Epoch: 0020 train cost=1.961569960 test cost=1.954860931\n",
            "Epoch: 0021 train cost=1.955118692 test cost=1.948297310\n",
            "Epoch: 0022 train cost=1.948766382 test cost=1.942062471\n",
            "Epoch: 0023 train cost=1.943004513 test cost=1.936123811\n",
            "Epoch: 0024 train cost=1.937135019 test cost=1.930460730\n",
            "Epoch: 0025 train cost=1.931803759 test cost=1.925039675\n",
            "Epoch: 0026 train cost=1.926716897 test cost=1.919830642\n",
            "Epoch: 0027 train cost=1.921655077 test cost=1.914828273\n",
            "Epoch: 0028 train cost=1.916846879 test cost=1.910010074\n",
            "Epoch: 0029 train cost=1.912354942 test cost=1.905377103\n",
            "Epoch: 0030 train cost=1.907873706 test cost=1.900894375\n",
            "Epoch: 0031 train cost=1.903676726 test cost=1.896562770\n",
            "Epoch: 0032 train cost=1.899277368 test cost=1.892373703\n",
            "Epoch: 0033 train cost=1.895417205 test cost=1.888319492\n",
            "Epoch: 0034 train cost=1.891568624 test cost=1.884386451\n",
            "Epoch: 0035 train cost=1.887803721 test cost=1.880562633\n",
            "Epoch: 0036 train cost=1.884060612 test cost=1.876853767\n",
            "Epoch: 0037 train cost=1.880632850 test cost=1.873247737\n",
            "Epoch: 0038 train cost=1.877162131 test cost=1.869743565\n",
            "Epoch: 0039 train cost=1.873571822 test cost=1.866331872\n",
            "Epoch: 0040 train cost=1.870173396 test cost=1.863002553\n",
            "Epoch: 0041 train cost=1.866803035 test cost=1.859759355\n",
            "Epoch: 0042 train cost=1.864134469 test cost=1.856599236\n",
            "Epoch: 0043 train cost=1.860774375 test cost=1.853520434\n",
            "Epoch: 0044 train cost=1.857890987 test cost=1.850507747\n",
            "Epoch: 0045 train cost=1.854895186 test cost=1.847571274\n",
            "Epoch: 0046 train cost=1.852226187 test cost=1.844698453\n",
            "Epoch: 0047 train cost=1.849296962 test cost=1.841887659\n",
            "Epoch: 0048 train cost=1.846570807 test cost=1.839146834\n",
            "Epoch: 0049 train cost=1.843970397 test cost=1.836457910\n",
            "Epoch: 0050 train cost=1.841330594 test cost=1.833834145\n",
            "Accuracy: 0.8631\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhYAAAEWCAYAAAA6r95OAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBQZeXEGwVAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LX_sAfKfi5uP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "9bbe3868-bd9d-4483-b2c1-233ca8c17cd8"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEMCAYAAADAqxFbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxOCccy6HJwbnnHM5PDE455zL8f+drHInj49elwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KDdtJGNLjNOx",
        "colab_type": "text"
      },
      "source": [
        "**Model 4: Input - Relu(BatchNormalization(512)) - Relu(BatchNormalization(128))- Sigmoid(output)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gD4tojuNi_We",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# https://www.tensorflow.org/api_docs/python/tf/nn/batch_normalization\n",
        "# https://r2rt.com/implementing-batch-normalization-in-tensorflow.html\n",
        "epsilon = 1e-3\n",
        "def multilayer_perceptron_batch(x, weights, biases):\n",
        "    # Use tf.matmul instead of \"*\" because tf.matmul can change it's dimensions on the fly (broadcast)\n",
        "    print( 'x:', x.get_shape(), 'W[h1]:', weights['h1'].get_shape(), 'b[h1]:', biases['b1'].get_shape())        \n",
        "    \n",
        "    ############################################################\n",
        "    # Hidden layer with Sigmoid activation and batch normalization\n",
        "    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) #(x*weights['h1']) + biases['b1']\n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/moments\n",
        "    # Calculate the mean and variance of x.\n",
        "    batch_mean_1, batch_var_1 = tf.nn.moments(layer_1,[0])\n",
        "    \n",
        "    scale_1 = tf.Variable(tf.ones([n_hidden_1]))\n",
        "    beta_1 = tf.Variable(tf.zeros([n_hidden_1]))\n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/batch_normalization\n",
        "    layer_1 = tf.nn.batch_normalization(layer_1, batch_mean_1, batch_var_1, beta_1, scale_1, epsilon)\n",
        "    layer_1 = tf.nn.relu(layer_1)\n",
        "    \n",
        "    print( 'layer_1:', layer_1.get_shape(), 'W[h2]:', weights['h2'].get_shape(), 'b[h2]:', biases['b2'].get_shape())        \n",
        "    \n",
        "    #####################################################################################\n",
        "    \n",
        "    # Hidden layer with Sigmoid activation and batch normalization\n",
        "    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) # (layer_1 * weights['h2']) + biases['b2'] \n",
        "    \n",
        "    # https://www.tensorflow.org/api_docs/python/tf/nn/moments\n",
        "    # Calculate the mean and variance of x.\n",
        "    batch_mean_2, batch_var_2 = tf.nn.moments(layer_2, [0])\n",
        "    \n",
        "    scale_2 = tf.Variable(tf.ones([n_hidden_2]))\n",
        "    beta_2 = tf.Variable(tf.zeros([n_hidden_2]))\n",
        "    \n",
        "    layer_2 = tf.nn.batch_normalization(layer_2, batch_mean_2, batch_var_2, beta_2, scale_2, epsilon)\n",
        "    layer_2 = tf.nn.relu(layer_2)\n",
        "    print( 'layer_2:', layer_2.get_shape(), 'W[out]:', weights['out'].get_shape(), 'b3:', biases['out'].get_shape())        \n",
        "    \n",
        "    ######################################################################################\n",
        "    \n",
        "    # output layer with Sigmoid activation \n",
        "    out_layer = tf.matmul(layer_2, weights['out']) + biases['out'] # (layer_2 * weights['out']) + biases['out']    \n",
        "    out_layer = tf.nn.sigmoid(out_layer)\n",
        "    print('out_layer:',out_layer.get_shape())\n",
        "\n",
        "    return out_layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ngc0W1Evj8hm",
        "colab_type": "text"
      },
      "source": [
        "Model 4 with Adam Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yNrjk8SojxL-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1305
        },
        "outputId": "2378bf14-d4d3-4317-a113-dfb2cbc7567b"
      },
      "source": [
        "# Since we are using sigmoid activations in hiden layers we will be using weights that are initalized as weights_sgd\n",
        "ybatch = multilayer_perceptron_batch(x, weights_sgd, biases)\n",
        "\n",
        "# https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits\n",
        "cost_batch = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = ybatch, labels = y_))\n",
        "# https://github.com/amitmac/Question-Answering/issues/2\n",
        "# there are many optimizers available: https://www.tensorflow.org/versions/r1.2/api_guides/python/train#Optimizers \n",
        "optimizer_batch_adam = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost_batch)\n",
        "optimizer_batch_sgdc = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost_batch)\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "\n",
        "            # here we use AdamOptimizer\n",
        "            _, c, w = sess.run([optimizer_batch_adam, cost_batch, weights_sgd], feed_dict={x: batch_xs, y_: batch_ys})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_batch, feed_dict={x: mnist.test.images, y_: mnist.test.labels})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax, np.arange(1.3, 1.8, step=0.04), \"input-Sigmoid(BN(512))-Sigmoid(BN(128))-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(ybatch,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels}))"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x: (?, 784) W[h1]: (784, 512) b[h1]: (512,)\n",
            "layer_1: (?, 512) W[h2]: (512, 128) b[h2]: (128,)\n",
            "layer_2: (?, 128) W[out]: (128, 10) b3: (10,)\n",
            "out_layer: (?, 10)\n",
            "Epoch: 0001 train cost=1.550719215 test cost=1.542217735\n",
            "Epoch: 0002 train cost=1.489568850 test cost=1.489876929\n",
            "Epoch: 0003 train cost=1.480976825 test cost=1.484142343\n",
            "Epoch: 0004 train cost=1.476471038 test cost=1.481534824\n",
            "Epoch: 0005 train cost=1.473359068 test cost=1.480400524\n",
            "Epoch: 0006 train cost=1.471591799 test cost=1.479522575\n",
            "Epoch: 0007 train cost=1.469827126 test cost=1.479425985\n",
            "Epoch: 0008 train cost=1.469395839 test cost=1.479432216\n",
            "Epoch: 0009 train cost=1.468494422 test cost=1.479062848\n",
            "Epoch: 0010 train cost=1.467538627 test cost=1.478531547\n",
            "Epoch: 0011 train cost=1.466986813 test cost=1.478505831\n",
            "Epoch: 0012 train cost=1.466439990 test cost=1.479148156\n",
            "Epoch: 0013 train cost=1.466226330 test cost=1.477945618\n",
            "Epoch: 0014 train cost=1.465848802 test cost=1.477917459\n",
            "Epoch: 0015 train cost=1.465735627 test cost=1.478504992\n",
            "Epoch: 0016 train cost=1.465073020 test cost=1.477868209\n",
            "Epoch: 0017 train cost=1.465052503 test cost=1.477412100\n",
            "Epoch: 0018 train cost=1.465070023 test cost=1.477995825\n",
            "Epoch: 0019 train cost=1.464737362 test cost=1.478032060\n",
            "Epoch: 0020 train cost=1.464288969 test cost=1.478059185\n",
            "Epoch: 0021 train cost=1.463902120 test cost=1.477882981\n",
            "Epoch: 0022 train cost=1.464500880 test cost=1.478306215\n",
            "Epoch: 0023 train cost=1.464070721 test cost=1.477606898\n",
            "Epoch: 0024 train cost=1.464390268 test cost=1.477500795\n",
            "Epoch: 0025 train cost=1.463789753 test cost=1.477504001\n",
            "Epoch: 0026 train cost=1.463788573 test cost=1.478262251\n",
            "Epoch: 0027 train cost=1.463544395 test cost=1.477340020\n",
            "Epoch: 0028 train cost=1.463941031 test cost=1.477783266\n",
            "Epoch: 0029 train cost=1.463889853 test cost=1.476743308\n",
            "Epoch: 0030 train cost=1.463449466 test cost=1.477141414\n",
            "Epoch: 0031 train cost=1.463613001 test cost=1.477607680\n",
            "Epoch: 0032 train cost=1.463216690 test cost=1.476648356\n",
            "Epoch: 0033 train cost=1.463206648 test cost=1.477098664\n",
            "Epoch: 0034 train cost=1.463110372 test cost=1.476896873\n",
            "Epoch: 0035 train cost=1.463508640 test cost=1.476922985\n",
            "Epoch: 0036 train cost=1.463451289 test cost=1.476774476\n",
            "Epoch: 0037 train cost=1.463264476 test cost=1.476577496\n",
            "Epoch: 0038 train cost=1.462875386 test cost=1.476270608\n",
            "Epoch: 0039 train cost=1.462872515 test cost=1.476748429\n",
            "Epoch: 0040 train cost=1.462956288 test cost=1.477083756\n",
            "Epoch: 0041 train cost=1.462985125 test cost=1.477537392\n",
            "Epoch: 0042 train cost=1.463028488 test cost=1.476642130\n",
            "Epoch: 0043 train cost=1.463218859 test cost=1.477335504\n",
            "Epoch: 0044 train cost=1.462874053 test cost=1.476949040\n",
            "Epoch: 0045 train cost=1.462870104 test cost=1.476706145\n",
            "Epoch: 0046 train cost=1.462765962 test cost=1.476642284\n",
            "Epoch: 0047 train cost=1.462704455 test cost=1.477265700\n",
            "Epoch: 0048 train cost=1.462653864 test cost=1.476547741\n",
            "Epoch: 0049 train cost=1.462506905 test cost=1.476667259\n",
            "Epoch: 0050 train cost=1.462426449 test cost=1.477040184\n",
            "Accuracy: 0.9824\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAdQAAAEWCAYAAADfB2bTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nzKwEnKGamZmVgDNUMzOzEvj/OiP/yrPGr7wAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EstqKN0EkxtX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "outputId": "fdca1cd2-9f83-42e7-e264-40906229c690"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "plt.show()"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEMCAYAAADAqxFbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyCTqqkNsnwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yLCLWbGv4XaQ",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "Model 5: Input - ReLu(512) - Dropout - ReLu(128)- Dropout -Sigmoid(output) "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6D-9kCmAreFr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Create model\n",
        "def multilayer_perceptron_dropout(x, weights, biases):\n",
        "    # Use tf.matmul instead of \"*\" because tf.matmul can change it's dimensions on the fly (broadcast)\n",
        "    print( 'x:', x.get_shape(), 'W[h1]:', weights['h1'].get_shape(), 'b[h1]:', biases['b1'].get_shape())        \n",
        "    \n",
        "    # we are adding a drop out layer after input layers with parameter keep_prob_input\n",
        "    # Hidden layer with ReLu activation\n",
        "    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) #(x*weights['h1']) + biases['b1']\n",
        "    layer_1 = tf.nn.relu(layer_1)\n",
        "    # we are adding a drop out layer after the first hidden layer with parameter keep_prob\n",
        "    layer_1_drop = tf.nn.dropout(layer_1, keep_prob)\n",
        "    \n",
        "    print( 'layer_1:', layer_1.get_shape(), 'W[h2]:', weights['h2'].get_shape(), 'b[h2]:', biases['b2'].get_shape())        \n",
        "    \n",
        "    # Hidden layer with ReLu activation\n",
        "    layer_2 = tf.add(tf.matmul(layer_1_drop, weights['h2']), biases['b2']) # (layer_1 * weights['h2']) + biases['b2'] \n",
        "    layer_2 = tf.nn.relu(layer_2)\n",
        "    \n",
        "    # we are adding a drop out layer after the first hidden layer with parameter keep_prob\n",
        "    layer_2_drop = tf.nn.dropout(layer_2, keep_prob)\n",
        "    print( 'layer_2:', layer_2.get_shape(), 'W[out]:', weights['out'].get_shape(), 'b3:', biases['out'].get_shape())        \n",
        "    \n",
        "    # Output layer with Sigmoid activation\n",
        "    out_layer = tf.matmul(layer_2_drop, weights['out']) + biases['out'] # (layer_2 * weights['out']) + biases['out']    \n",
        "    out_layer = tf.nn.sigmoid(out_layer)\n",
        "    print('out_layer:',out_layer.get_shape())\n",
        "\n",
        "    return out_layer"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "91ITUaBT5LHr",
        "colab_type": "text"
      },
      "source": [
        "Model 5 with Adam Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pPD4svzu5Eu7",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1380
        },
        "outputId": "4403ebd1-ad49-4c61-a5e9-7a28d3d085e0"
      },
      "source": [
        "# Since we are using Relu activations in hiden layers we will be using weights that are initalized as weights_relu\n",
        "ydrop = multilayer_perceptron_dropout(x, weights_relu, biases)\n",
        "\n",
        "# https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits\n",
        "cost_drop = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = ydrop, labels = y_))\n",
        "# https://github.com/amitmac/Question-Answering/issues/2\n",
        "# there are many optimizers available: https://www.tensorflow.org/versions/r1.2/api_guides/python/train#Optimizers \n",
        "optimizer_drop_adam = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost_drop)\n",
        "optimizer_drop_sgdc = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost_drop)\n",
        "\n",
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "            \n",
        "            # here we use AdamOptimizer\n",
        "            _, c, w = sess.run([optimizer_drop_adam, cost_drop, weights_relu], feed_dict={x: batch_xs, y_: batch_ys, keep_prob: 0.5})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_drop, feed_dict={x: mnist.test.images, y_: mnist.test.labels,  keep_prob: 1.0})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax,np.arange(1, 1.8, step=0.05), \"input-ReLu(512)-Dropout-ReLu(128)-Dropout-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax,np.arange(1, 1.8, step=0.05), \"input-ReLu(512)-Dropout-ReLu(128)-Dropout-Sigmoid(output)-AdamOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(ydrop,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0 }))"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x: (?, 784) W[h1]: (784, 512) b[h1]: (512,)\n",
            "WARNING:tensorflow:From <ipython-input-48-5214f7a2eda2>:10: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "layer_1: (?, 512) W[h2]: (512, 128) b[h2]: (128,)\n",
            "layer_2: (?, 128) W[out]: (128, 10) b3: (10,)\n",
            "out_layer: (?, 10)\n",
            "Epoch: 0001 train cost=1.681832925 test cost=1.606635197\n",
            "Epoch: 0002 train cost=1.551047493 test cost=1.517656306\n",
            "Epoch: 0003 train cost=1.526984276 test cost=1.504301678\n",
            "Epoch: 0004 train cost=1.516647001 test cost=1.497379170\n",
            "Epoch: 0005 train cost=1.508817648 test cost=1.493293790\n",
            "Epoch: 0006 train cost=1.505042977 test cost=1.490217356\n",
            "Epoch: 0007 train cost=1.501550822 test cost=1.488461393\n",
            "Epoch: 0008 train cost=1.499076786 test cost=1.486802942\n",
            "Epoch: 0009 train cost=1.495452141 test cost=1.485375289\n",
            "Epoch: 0010 train cost=1.493345639 test cost=1.484503217\n",
            "Epoch: 0011 train cost=1.491519288 test cost=1.484214500\n",
            "Epoch: 0012 train cost=1.490384412 test cost=1.483649907\n",
            "Epoch: 0013 train cost=1.489159080 test cost=1.482995322\n",
            "Epoch: 0014 train cost=1.487652125 test cost=1.482813195\n",
            "Epoch: 0015 train cost=1.486633029 test cost=1.482033671\n",
            "Epoch: 0016 train cost=1.485923679 test cost=1.482227993\n",
            "Epoch: 0017 train cost=1.485342925 test cost=1.481754916\n",
            "Epoch: 0018 train cost=1.484734865 test cost=1.481461066\n",
            "Epoch: 0019 train cost=1.484194670 test cost=1.480634614\n",
            "Epoch: 0020 train cost=1.484140284 test cost=1.480624328\n",
            "Epoch: 0021 train cost=1.482856067 test cost=1.480347208\n",
            "Epoch: 0022 train cost=1.482969127 test cost=1.480200718\n",
            "Epoch: 0023 train cost=1.482042490 test cost=1.480244198\n",
            "Epoch: 0024 train cost=1.482244338 test cost=1.479340020\n",
            "Epoch: 0025 train cost=1.481157897 test cost=1.479613139\n",
            "Epoch: 0026 train cost=1.481073100 test cost=1.479996362\n",
            "Epoch: 0027 train cost=1.480574587 test cost=1.479978995\n",
            "Epoch: 0028 train cost=1.480466675 test cost=1.480051087\n",
            "Epoch: 0029 train cost=1.480719312 test cost=1.478966502\n",
            "Epoch: 0030 train cost=1.480820225 test cost=1.479554286\n",
            "Epoch: 0031 train cost=1.480312479 test cost=1.479232365\n",
            "Epoch: 0032 train cost=1.480032930 test cost=1.478901256\n",
            "Epoch: 0033 train cost=1.479830554 test cost=1.478768641\n",
            "Epoch: 0034 train cost=1.479530471 test cost=1.479310688\n",
            "Epoch: 0035 train cost=1.479219449 test cost=1.479274155\n",
            "Epoch: 0036 train cost=1.479710231 test cost=1.479170908\n",
            "Epoch: 0037 train cost=1.478987146 test cost=1.479316568\n",
            "Epoch: 0038 train cost=1.478956976 test cost=1.479022618\n",
            "Epoch: 0039 train cost=1.478688333 test cost=1.479144982\n",
            "Epoch: 0040 train cost=1.479036336 test cost=1.478652251\n",
            "Epoch: 0041 train cost=1.478400348 test cost=1.478876143\n",
            "Epoch: 0042 train cost=1.478042485 test cost=1.478665199\n",
            "Epoch: 0043 train cost=1.477481085 test cost=1.478196388\n",
            "Epoch: 0044 train cost=1.478215101 test cost=1.477784476\n",
            "Epoch: 0045 train cost=1.477765940 test cost=1.478228672\n",
            "Epoch: 0046 train cost=1.477335376 test cost=1.477820701\n",
            "Epoch: 0047 train cost=1.477252278 test cost=1.477749380\n",
            "Epoch: 0048 train cost=1.477474598 test cost=1.478463836\n",
            "Epoch: 0049 train cost=1.478009496 test cost=1.478034486\n",
            "Epoch: 0050 train cost=1.476928517 test cost=1.478232527\n",
            "Accuracy: 0.9791\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAEWCAYAAAC39OTrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDshmZmZNwAHZzMysCTggm5mZNYH/DwasVmXh7l61AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RChM12gT5P28",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        },
        "outputId": "872dd68b-5d64-4f97-d5a8-2012be5d8d7f"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "# plt.show()"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 0, 'Output Layer ')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 50
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEMCAYAAADNtWEcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KVZL1VW35xB_",
        "colab_type": "text"
      },
      "source": [
        "Model 5 with Gradient Descent Optimizer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2u-p3uej5t4m",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1232
        },
        "outputId": "c5b2ab4d-f6e8-4656-f55a-87a17cdd0ac4"
      },
      "source": [
        "with tf.Session() as sess:\n",
        "    tf.global_variables_initializer().run()\n",
        "    fig,ax = plt.subplots(1,1)\n",
        "    ax.set_xlabel('epoch') ; ax.set_ylabel('Soft Max Cross Entropy loss')\n",
        "    xs, ytrs, ytes = [], [], []\n",
        "    for epoch in range(training_epochs):\n",
        "        train_avg_cost = 0.\n",
        "        test_avg_cost = 0.\n",
        "        total_batch = int(mnist.train.num_examples/batch_size)\n",
        "\n",
        "        # Loop over all batches\n",
        "        for i in range(total_batch):\n",
        "            batch_xs, batch_ys = mnist.train.next_batch(batch_size)\n",
        "            \n",
        "            # here we use GradientDescentOptimizer\n",
        "            _, c, w = sess.run([optimizer_drop_sgdc, cost_drop, weights_relu], feed_dict={x: batch_xs, y_: batch_ys, keep_prob: 0.5})\n",
        "            train_avg_cost += c / total_batch\n",
        "            c = sess.run(cost_drop, feed_dict={x: mnist.test.images, y_: mnist.test.labels,  keep_prob: 1.0})\n",
        "            test_avg_cost += c / total_batch\n",
        "\n",
        "        xs.append(epoch)\n",
        "        ytrs.append(train_avg_cost)\n",
        "        ytes.append(test_avg_cost)\n",
        "        plt_dynamic(xs, ytrs, ytes, ax,np.arange(1.6, 2.4, step=0.05), \"input-ReLu(512)-Dropout-ReLu(128)-Dropout-Sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "        if epoch%display_step == 0:\n",
        "            print(\"Epoch:\", '%04d' % (epoch+1), \"train cost={:.9f}\".format(train_avg_cost), \"test cost={:.9f}\".format(test_avg_cost))\n",
        "\n",
        "    # plot final results\n",
        "    plt_dynamic(xs, ytrs, ytes, ax,np.arange(1.6, 2.4, step=0.05), \"input-ReLu(512)-Dropout-ReLu(128)-Dropout-Sigmoid(output)-GradientDescentOptimizer\")\n",
        "\n",
        "    # we are calculating the final accuracy on the test data\n",
        "    correct_prediction = tf.equal(tf.argmax(ydrop,1), tf.argmax(y_,1))\n",
        "    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
        "    print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0 }))"
      ],
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch: 0001 train cost=2.367956066 test cost=2.329291472\n",
            "Epoch: 0002 train cost=2.349387010 test cost=2.294113149\n",
            "Epoch: 0003 train cost=2.330561284 test cost=2.260481291\n",
            "Epoch: 0004 train cost=2.314644326 test cost=2.229324016\n",
            "Epoch: 0005 train cost=2.299356648 test cost=2.202690001\n",
            "Epoch: 0006 train cost=2.284356543 test cost=2.181132024\n",
            "Epoch: 0007 train cost=2.268288464 test cost=2.160883063\n",
            "Epoch: 0008 train cost=2.254606537 test cost=2.138644694\n",
            "Epoch: 0009 train cost=2.239533172 test cost=2.113463464\n",
            "Epoch: 0010 train cost=2.227186948 test cost=2.085177789\n",
            "Epoch: 0011 train cost=2.209031092 test cost=2.054540972\n",
            "Epoch: 0012 train cost=2.195728822 test cost=2.023238482\n",
            "Epoch: 0013 train cost=2.175764040 test cost=1.992266788\n",
            "Epoch: 0014 train cost=2.160595560 test cost=1.963339296\n",
            "Epoch: 0015 train cost=2.142267095 test cost=1.938043228\n",
            "Epoch: 0016 train cost=2.125118529 test cost=1.916398488\n",
            "Epoch: 0017 train cost=2.108566491 test cost=1.896317577\n",
            "Epoch: 0018 train cost=2.095649126 test cost=1.879818382\n",
            "Epoch: 0019 train cost=2.078838087 test cost=1.865015228\n",
            "Epoch: 0020 train cost=2.064751520 test cost=1.852745039\n",
            "Epoch: 0021 train cost=2.052643522 test cost=1.841662188\n",
            "Epoch: 0022 train cost=2.038157456 test cost=1.831312898\n",
            "Epoch: 0023 train cost=2.026730894 test cost=1.822345859\n",
            "Epoch: 0024 train cost=2.017024798 test cost=1.814359866\n",
            "Epoch: 0025 train cost=2.005036580 test cost=1.806824874\n",
            "Epoch: 0026 train cost=1.993808699 test cost=1.799933447\n",
            "Epoch: 0027 train cost=1.981616905 test cost=1.793593432\n",
            "Epoch: 0028 train cost=1.974486065 test cost=1.787660274\n",
            "Epoch: 0029 train cost=1.963341850 test cost=1.782237238\n",
            "Epoch: 0030 train cost=1.957479723 test cost=1.777416952\n",
            "Epoch: 0031 train cost=1.948664274 test cost=1.772978413\n",
            "Epoch: 0032 train cost=1.939965757 test cost=1.768793451\n",
            "Epoch: 0033 train cost=1.933158220 test cost=1.764748280\n",
            "Epoch: 0034 train cost=1.924310863 test cost=1.761080598\n",
            "Epoch: 0035 train cost=1.919065479 test cost=1.757603109\n",
            "Epoch: 0036 train cost=1.912751099 test cost=1.754348021\n",
            "Epoch: 0037 train cost=1.905422830 test cost=1.751239256\n",
            "Epoch: 0038 train cost=1.902098959 test cost=1.748623948\n",
            "Epoch: 0039 train cost=1.895192638 test cost=1.745797869\n",
            "Epoch: 0040 train cost=1.891789072 test cost=1.742948265\n",
            "Epoch: 0041 train cost=1.885219745 test cost=1.740363558\n",
            "Epoch: 0042 train cost=1.880090311 test cost=1.738020679\n",
            "Epoch: 0043 train cost=1.873409373 test cost=1.735685815\n",
            "Epoch: 0044 train cost=1.872536089 test cost=1.733344485\n",
            "Epoch: 0045 train cost=1.865687063 test cost=1.730953564\n",
            "Epoch: 0046 train cost=1.861622271 test cost=1.728924230\n",
            "Epoch: 0047 train cost=1.858306549 test cost=1.726784129\n",
            "Epoch: 0048 train cost=1.852631298 test cost=1.724687271\n",
            "Epoch: 0049 train cost=1.850821122 test cost=1.722752868\n",
            "Epoch: 0050 train cost=1.847244602 test cost=1.720746724\n",
            "Accuracy: 0.7693\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAicAAAEWCAYAAAC9njdIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nbULdH1o7AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OGo0Y8rr9JHo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 303
        },
        "outputId": "dcbdf9dd-7291-4abb-f4a1-44a0813cc794"
      },
      "source": [
        "h1_w = w['h1'].flatten().reshape(-1,1)\n",
        "h2_w = w['h2'].flatten().reshape(-1,1)\n",
        "out_w = w['out'].flatten().reshape(-1,1)\n",
        "\n",
        "\n",
        "fig = plt.figure()\n",
        "plt.subplot(1, 3, 1)\n",
        "plt.title(\"Weight matrix\")\n",
        "ax = sns.violinplot(y=h1_w,color='b')\n",
        "plt.xlabel('Hidden Layer 1')\n",
        "\n",
        "plt.subplot(1, 3, 2)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=h2_w, color='r')\n",
        "plt.xlabel('Hidden Layer 2 ')\n",
        "\n",
        "plt.subplot(1, 3, 3)\n",
        "plt.title(\"Weight matrix \")\n",
        "ax = sns.violinplot(y=out_w,color='y')\n",
        "plt.xlabel('Output Layer ')\n",
        "# plt.show()"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 0, 'Output Layer ')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 52
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAEMCAYAAAAs8rYIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}