Implementation of the regularized least-squares algorithm in MATLAB

The regularized least-squares algorithm is used in regression for finding machine learning models. This implementation of the RLM uses a polynomial curve fitting problem to analyze the performance over the ERM.

Features

• RLM analysis for w = 1, 2,..., 30 with different regularizers;
• ERM analysis for w = 1, 2,..., 30;
• k-fold cross validation over ERM;
• Empirical loss comparison;

Dataset

An 1-imput dataset with 100 data points is used for training the models. This dataset was originally proposed by Dr. Ruth Urner on one of her assignments for ML.

Documentation

This implementation is part of these two articles published on the Towards Data Science blog. I encourage you to go over these two readings.