SaniyaAbushakimova / A-Study-of-Sigmoid-based-Multi-class-Logistic-Regression

Investigated how multi-class logistic regression would perform if the activation function was changed from softmax to sigmoid. It included mathematical analysis and empirical evaluation, such as rewriting the model from scratch. Tech: Python (Scikit-Learn, Pandas)

Purpose

• Investigate how sigmoid-based multi-class logistic regression would perform compared to softmax_based one.

Content

• Final Project/Final_project_report.pdf - Report (AAAI format) that contains description and results of the work done.
• Final Project/Final_project_solution.ipynb - Jupyther Notebook with Data exploration, training procedures and experiments.
• Final Project/Preprocessing.ipynb - Jupyter Notebook with data preprocessing.