Overview

Description

CALENDAR DESCRIPTION:

Statistical foundation for machine learning algorithms, emphasizing bias-variance tradeoff. Students will learn principles for choosing effective methods and tailoring them to fit a given learning problem. Potential topics include probabilistic graphical models, maximum likelihood estimation, latent variables and the EM algorithm, convex optimization, and variational and sampling-based methods.

COURSE DETAILS:

Why we're offering this course: Machine learning now plays a central role in hundreds of fields. All learning methods have common underpinnings based on probability and statistics, but these are not widely understood. This course aims to give students a probabilistic foundation for machine learning and an understanding of probabilistic machine learning methods. ------------------- After you complete this course, you will be able to: * Design and implement probabilistic machine learning techniques including: probabilistic graphical models, discrete and continuous distributions, maximum likelihood estimation, EM algorithm inference, sampling-based inference methods, MCMC, variational inference. * Understand concepts such as: prior, posterior, likelihood, overfitting, bias-variance tradeoff, likelihood, regularization. * Choose between probabilistic machine learning methods, understand what aspects of a data set influence machine learning performance. and foresee which will perform best ------------------- Who should take this course? This course is intended for graduate students with an interest in machine learning or big data. You should take this course if: * You intend to use or develop machine learning in your research or work. * You are interested in fields where probabilisitic machine learning is important including: vision, natural language processing/understanding, medical imaging, robotics, smart cities. * You want to understand statistical behavior of machine learning methods at a deep level. ------------------- Prerequisites: No official prerequisites. However, the course assumes basic knowledge of machine learning (e.g. CMPT 726), probability (e.g. STAT 270) and linear algebra (e.g. MATH 240). The course is open to advanced undergraduates with permission.

Topics

• Probabilistic graphical models
• Discrete and continuous distributions
• Maximum likelihood estimation
• EM algorithm
• Bayesian probability

Grading