-
Multiclass Classification - Overview
Learn how One-vs-All and One-vs-One extend binary classification to multiclass problems, their key differences, and best use cases.
-
Gaussian Regression - A Bayesian Approach to Linear Regression
This guide explores Gaussian regression, deriving its closed-form posterior, linking MAP estimation to ridge regression, and explaining predictive uncertainty for Bayesian inference.
-
My Understanding of "Efficient Algorithms for Online Decision Problems" Paper
A breakdown of Follow the Perturbed Leader (FPL) from Kalai & Vempala’s (2005) paper, "Efficient Algorithms for Online Decision Problems." This blog explores how FPL improves online decision-making, minimizes regret, and extends to structured problems like shortest paths and adaptive Huffman coding.
-
Follow the Leader (FL) and Follow the Perturbed Leader (FPL) in Online Learning
Discover how Follow the Leader (FL) and Follow the Perturbed Leader (FPL) work in online learning, their mathematical foundations, and how perturbations help achieve better stability and regret bounds.
-
Bayesian Conditional Models
Learn how Bayesian conditional models leverage prior knowledge, posterior updates, and predictive distributions to make principled, uncertainty-aware predictions in machine learning.