<aside> 📣 Announcements: Alternative talk for (May 24): ****https://www.youtube.com/watch?v=8eW4ZTfMLOI Google form (May 24): https://forms.gle/je3R6QCnBowGZcCs6 ****Alternative talk for (May 31): ****https://www.youtube.com/watch?v=JMO1YSSQKfY Google form (May 31): https://forms.gle/7BxrcYX9CE7Fn7PY7

</aside>

Overview

This course presents a signal processing perspective on machine learning. The goal is to employ a systematic framework grounded in probabilistic reasoning and optimization, in order to gain a fundamental understanding of the “best” approaches out there for topics such as unsupervised and supervised learning, and generative models. First half of the class will be more “classical” ML methods, and the second half will be cover more “modern” ML approaches based on neural networks.

Prerequisites: Great comfort with probability (ECE 235) and linear algebra, or the mathematical sophistication required to pick up the relevant background quickly from the resources that we will link to. Familiarity with python or commitment to significant up-front effort to pick it up.

Learning Outcomes

  1. Students will be able to construct and analyze a mathematical model for a given machine learning problem.
  2. Students will be able to explain core concepts in classical and modern machine learning techniques. Classical methods include logistic regression, support vector machines, clustering, etc. Modern methods include feedforward neural networks, graph neural networks, generative neural networks, etc.
  3. Students will be able to develop Python codes for small machine learning tasks (both classical and modern).

Evaluations