Machine Learning Project 1

First Project of the Machine Learning course

Repository Structure

• code - Folder containing the implementation of the Deep Frank-Wolfe Algorithm
  • implementations.py - Implementation of all required optimization algorithms.
  • proj1_helpers.py - Utilities for training.
  • proj1_input_manipulation.py - Preprocessing utilities.
  • proj1_linear_model.py - GD and SGD utilities.
  • proj1_linear_model.py - GD and SGD utilities.
  • proj1_logistic.py - Logistic regression utilities.
  • proj1_ridge_regression.py - Ridge regression utilities.
  • run.py - Reproducibility of best model.
  • run_grad.ipynb - Reproducibility of best GD results.
  • run_least_squares.ipynb - Reproducibility of best LS results.
  • run_logistic.ipynb - Reproducibility of best model.
  • run_regularized_logistic.ipynb - Reproducibility of best regularized logistic regression results.
  • run_ridge.ipynb - Reproducibility of best ridge regression results.
  • run_stochastic_grad.ipynb - Reproducibility of best SGD results.
• data - Contains a zip with the original dataset.
• report - This folder contains the report of the obtained results
  • report.pdf - Report pdf file
• requirements.txt - Requirements text file

Installation

To clone the following repository, please run:
git clone --recursive https://github.com/johnmavro/Machine-Learning-P1.git

Requirements

Requirements for the needed packages are available in requirements.txt. To install the needed packages, please run:
pip install -r requirements.txt

Reproducibility of the results

The notebooks for reproducibility of the best training results can be found in the code folder. Please refer to the repository description above for detailed instructions.

Report

The report in pdf format can be found in the folder report.

Authors

• Federico Betti
• Ioannis Mavrothalassitis
• Luca Rossi