README.md

Gradient descent and logistic regression

A modular gradient-based optimization implementation with alternative learning-rate policies, objective modules, and a logistic regression estimator.

Implementation

Component	File
Iterative gradient descent and stopping criteria	gradient_descent.py
Fixed and exponential learning-rate schedules	learning_rate.py
L1, L2, logistic and regularized modules	modules.py
Logistic regression estimator	logistic_regression.py
Gradient trajectory and learning-rate experiments	gradient_descent_investigation.py
Model and loss interfaces	base_module.py, base_estimator.py, loss_functions.py

Results

The implementation supports different returned solutions (last, best, average) and callback hooks for inspecting progress during optimization.

Environment: NumPy, scikit-learn, pandas and Plotly; Kaleido may be needed to export figures. Parts of gradient_descent_investigation.py expect the external SAheart.data dataset. The original module-level imports and exercise structure are preserved.