Please bring your Eventbrite ticket (print or digital) and your student ID (if you selected a student ticket) to your first lecture, to sign in and collect your name badge. Without the correct documentation you will not be permitted to attend. Please keep your name badge for all future lectures as confirmation of signing in. Without your name badge you will not be permitted to attend a lecture.

Registered participants can access the lecture slides library here, with the username and access code provided to them via email.

This Fields Institute graduate course will survey a selection of topics in the mathematics of machine learning, such as deep learning, automatic differentiation, non-convex optimization, probablistic modelling, stochastic variations, compressibility, probablistic inference, generative models, adversarial robustness, reinforcement learning and statistical learning theory. Applications to problems in medicine, finance, and manufacturing processes will also be discussed. The course will begin with several weeks of introductory material presented by members of the computer science department (Professors David Duvenaud and Roger Grosse), before transitioning to a series of short (3 hours) modules presented by visitors to the Fields Institute and Vector Institute staff on specific topics and applications.

Prerequisites

Undergraduate level probability, statistics, multivariable Calculus, linear algebra

Evaluation

Credit/no credit based on end-of-term presentations. Presentations may be based on a topic of the participants choosing that is related to course material.
Date for the presentation: mid/late April. Length: TBA
To arrange obtaining credit for this course please contact the Chair of your Graduate Department. Fields can provide the necessary supporting documentation to your Graduate Department on request.
Please direct queries to Brittany Camp the Fields-CQAM Liaison Officer.

Teaching Assistant Information

Name: Tristan Milne
Contact:

Office Hours: Thursdays 3pm - 4pm
Office Location: Fields Institute

Recommended Reading

Introductory machine learning references:
Neural Networks and Deep Learning; Michael Nielsen.
Deep Learning; Ian Goodfellow, Yoshua Bengio, and Aaron Courville

Lecture 12. Understanding Machine Learning by Shai Shalev-Shwartz and Shai Ben-David. Chapters 2-6, 26, 28 give the fundamentals of empirical risk minimization. Chapters 12-14, 26 give the fundamentals of regularized loss minimization for convex problems.

Topics covered:

I. Introduction to Deep Learning (DD & RG)

II. Automatic Differentiation (DD & RG)

III. Optimization (DD & RG)

IV. Probabilistic Modeling (DD & RG)

V. Stochastic Variations (DD & RG)

VI. Probabilistic Inference and Generative Models (AM)

VII. Reinforcement Learning (AF)

VIII. Regularization and Adversarial Robustness (AO)

IX. Latent variable analysis with application in advanced manufacturing processes (CD)

X. Machine Learning in Finance (LW)

XI. Deep Learning (GH)

XII. Unsupervised Learning with Application in Medicine (HW)

XIII. Statistical Learning Theory and Compressibility (KD)

XIV. Non-convex Optimization in ML (ME)

XV. Machine Learning in Health (DU)