MDL Complexity computations and experiments from the paper "Revisiting complexity and the bias-variance tradeoff".

Machine-Learning-Regression

L2 regularization, or Ridge regression, is a technique to prevent overfitting in machine learning by adding a penalty proportional to the sum of squared weights to the loss function. It forces weights to be small but rarely zero, resulting in a smoother, more stable model. Solver

Underfitting and overfitting are critical concepts in machine learning, particularly when using Polynomial Regression to model data. Polynomial regression allows a model to learn non-linear relationships by increasing the polynomial degree (e.g. ), making it highly susceptible to both underfitting (too simple) and overfitting (too complex).Solver

This repository includes some detailed proofs of "Bias Variance Decomposition for KL Divergence".

The projects are part of the graduate-level course CSE-574 : Introduction to Machine Learning [Spring 2019 @ UB_SUNY] . . . Course Instructor : Mingchen Gao (https://cse.buffalo.edu/~mgao8/)

This is a simple python example to demonstrate bias variance

The Bias-Variance Tradeoff Visualization project provides an interactive tool to understand the bias-variance tradeoff in machine learning models. It visually demonstrates how different models perform on training and validation datasets, helping users grasp the concepts of overfitting and underfitting.

Bias variance experiment from Learning from Data. Problem 2.24, p. 75.

Nine diagnostic tools for detecting and understanding overfitting in scikit-learn models — polynomial overfitting, learning curves, validation curves, bias-variance decomposition, regularisation sweeps, data leakage detection, and more. Companion code for the ML Diagnostics Mastery series.

Performing polynomial regression of varying degrees on data affected by white and Poisson noise, evaluating the model performance based on MSE loss and the bias-variance trade-off.

Explanation of the Bias Variance Tradeoff in Machine Learning

Interactive ML experimentation platform for understanding model behavior and statistical decision-making (A/B testing, bias-variance, learning curves).

This project focuses on developing and training supervised learning models for prediction and classification tasks, covering linear and logistic regression (using NumPy & scikit-learn), neural networks (with TensorFlow) for binary and multi-class classification, and decision trees along with ensemble methods like random forests and boosted trees

This project plots the variation of Bias, Variance and Generalization Error with k for k-NN on a fixed function

An R package and Shiny app for exploring the bias-variance tradeoff in polynomial and k-NN regression via Monte Carlo simulation.

This repository contains a generalized regression analysis problem solved from scratch, using only the Numpy library.

Estimating the parametric complexity (minimum description length) of binary classifiers.

Exploration of model performance, bias-variance tradeoffs, and dataset effects on classification accuracy with fully reproducible simulated data. This project was done in Python.