Added some old code to the paste-bin. Tested I/O-extraction with tupl…

…es in for-loop on the workbench. Added new method to the SNN which will extract neo. pseudopat. I.

Process Files/PasteBin.py

@@ -46,4 +46,25 @@
     #             if np.random.random() < self.MF:
     #                 self.update_dg_ca3_weights_value(random_dg_neuron_index, ca3_neuron_index, np.random.random())
     #             else:
-    #                 self.update_dg_ca3_weights_value(random_dg_neuron_index, ca3_neuron_index, 0.0)
+    #                 self.update_dg_ca3_weights_value(random_dg_neuron_index, ca3_neuron_index, 0.0)
+
+
+# def hpc_learn_patterns_iterations_hardcoded_wrapper(hpc, patterns):
+#     print "Commencing learning of", len(patterns), "I/O patterns."
+#     time_start_overall = time.time()
+#     iter_ctr = 0
+#     while iter_ctr < 2:
+#         p_ctr = 0
+#         for [input_pattern, output_pattern] in patterns:
+#             # Neuronal turnover, setting input and output in the hpc network.
+#             hpc.setup_pattern(input_pattern, output_pattern)
+#
+#             # one iteration of learning using Hebbian learning
+#             hpc.learn()
+#             p_ctr += 1
+#
+#         iter_ctr += 1
+#     time_stop_overall = time.time()
+#
+#     print "Learned", len(patterns), "pattern-associations in ", iter_ctr, "iterations, which took" "{:7.3f}". \
+#         format(time_stop_overall-time_start_overall), "seconds."

Process Files/TheanoWorkbench.py

@@ -18,11 +18,11 @@
 # # print binomial_vector_f([m])
 # # print uniform_f(2, 10)
 #
-hpc = HPC([49, 240, 1600, 480, 49],
-          0.67, 0.25, 0.04,  # connection rates: (in_ec, ec_dg, dg_ca3)
-          0.10, 0.01, 0.04,  # firing rates: (ec, dg, ca3)
-          0.7, 1.0, 0.1, 0.5,  # gamma, epsilon, nu, turnover rate
-          0.10, 0.95, 0.8, 2.0)  # k_m, k_r, a_i, alpha, alpha is 2 in 4.1.1
+# hpc = HPC([49, 240, 1600, 480, 49],
+#           0.67, 0.25, 0.04,  # connection rates: (in_ec, ec_dg, dg_ca3)
+#           0.10, 0.01, 0.04,  # firing rates: (ec, dg, ca3)
+#           0.7, 1.0, 0.1, 0.5,  # gamma, epsilon, nu, turnover rate
+#           0.10, 0.95, 0.8, 2.0)  # k_m, k_r, a_i, alpha, alpha is 2 in 4.1.1
 
 # test = np.asarray([np.arange(5), np.arange(5)*2])
 # shared_test = theano.shared(name='shared_test', value=test.astype(theano.config.floatX), borrow=True)
@@ -42,3 +42,9 @@
 # print "after row replacement\n:", shared_test.get_value()
 # replace_column(2, np.asarray([42, 42], dtype=np.float32))
 # print "after column update:\n", shared_test.get_value()
+
+test_list = [[2, 4]] * 3
+test_list.append([13, 9])
+for [i, o] in test_list:
+    print "i:", i
+    print "o:", o

SimpleNeocorticalNetwork.py

@@ -1,5 +1,6 @@
 import theano
 import theano.tensor as T
+from theano.tensor.shared_randomstreams import RandomStreams
 import numpy as np
 
 theano.config.floatX = 'float32'
@@ -90,6 +91,13 @@ def train(self, IOPairs):
                                 self.prev_delta_W_in_h.get_value(return_internal_type=True),
                                 self.prev_delta_W_h_out.get_value(return_internal_type=True))
 
+    def get_pseudopattern_I(self):
+        # random input
+        rng = RandomStreams()
+
+        # get output
+        # return [I, O]
+
     def print_layers(self):
         print "\nPrinting layer activation values:"
         print "input:\t", self._in.get_value()