Logistic-Regression

Description

UNM CS 529 Project 2: Creation of logistic regression classifier from scratch.

Instructions for Use

Install Dependencies

Create a Python virtual environment

python -m venv YOURVENV

Activate the environment by running this command on Windows...

YOURENV/Scripts/activate

or this command on Linux/MacOS

source ./YOURENV/bin/activate

Install required dependecies with the command below (all platforms):

pip install -r requirements.txt

Train Logistic Regression

1. Specify values for the variables below in utils/consts.py:

   • training_data_path: the path to your training data directory
   • testing_data_path: the path to your testing (kaggle) data directory

2. Run python -m training.train_logistic_regression from the top level directory.

3. The following files will be generated:

• A file will be generated containing the trained model and saved in the models/ directory.
• A file will be generated containing kaggle predictions and saved as kaggle_predictions.csv in the top level directory.

Code Manifest

File Name	Description
training/train_logistic_regression.py	This file contains the implementation of logistic regression and gradient descent.
training/library_models.py	This file contains our implementation of training and validation of scikit-learn ML models for comparison to our logistic regression implementation.
plots/convergence_plots.py	This file contains a script for generating convergence rate plots.
utils/process_audio_data.py	This file contains our feature extraction and transformation implementation.
utils/consts.py	This file has constants used throughout the library.
utils/file_utils.py	This file contains utility functions for working with files.
validation/validate.py	This file contains our function to generate kaggle prediction CSV files.

Developer Contributions

Prasanth Reddy Guvvala

• Implemented gradient descent algorithm.
• Implemented statistical feature transformations.
• Wrote script to generate confusion matrix.

Thomas Fisher

• Implemented combined feature extraction algorithm.
• Implemented scikit-learn library functions.
• Wrote script to plot convergence rate.

kaggle Submission

Leaderboard position 5 achieved with accuracy 0.71 on April 8th (team name: Fisher & Guvvala).