Probabilistic Machine Learning

Course Description

This is an introductory course to topics in machine learning studied from a probability theory viewpoint. The course covers an overview of basic probability, inference and estimation, regression algorithms, Markov chains, inference for Markov models and the EM algorithm, Markov decision process, and reinforcement learning. In addition to covering mathematical and algorithmic details, the course includes several hands-on projects to implement the machine learning algorithms.

References (not needed but useful for further studies)

• Kevin Murphy, Machine Learning: A Probabilistic Perspective, 2012.
• Jean Walrand, Probability in Electrical Engineering and Computer Science: An Application-Driven Course, 2014.

Tentative Course Outline

1. Review of Probability

Probability space, random variables, probability distribution, limit theorems (3 lectures)

2. Bayesian Learning

MAP estimation, maximum likelihood estimation, linear regression, ridge regression,logistic regression (3 lectures)

3. Supervised ML Algorithms

Empirical Risk Minimization, Gradient Descent, Stochastic Gradient Descent, Generalization Error (3 lectures)

4. Unsupervised Learning

Clustering, K-means Algorithm, Expectation Maximization (EM) algorithm (3 lectures)

5. Markov Models

Markov Chains, Hidden Markov Models, Markov Decision Process (MDP) (3 lectures)

6. Reinforcement Learning

Dynamic Programming, Value Iteration, Q Learning, Introduction to Deep Reinforcement Learning (3 lectures)