Machine Learning (DSGA-1003)

Course Description

The course covers general machine learning methods including generalized linear models, graphical models, causal inference, and reinforcement learning. There will be homework assignments along with a final project. Reading responses will be required every week along with a scribe of one lecture.

Lecture Location

Tuesday 4:55-6:35pm; Location: 238 Thompson St (GCASL) Room C95

Reading Assignments

• Introduction and Overview (1/23/24)
• Supervised Learning (1/30/24)
  • [Murphy, Book 1, Ch 1.1-1.2: Supervised Learning]
  • [Murphy, Book 1, Ch 8.2.1-8.2.2: First-order Methods]
  • [Murphy, Book 1, Ch 8.4.1-8.4.4: Stochastic Gradient Descent]
• A Bayesian Interlude (2/6/24)
  • [Murphy, Book 1, Ch 11.7: Bayesian linear regression]
• Regularization + Generalized Linear Models (2/13/24)
  • [Murphy, Book 1, Chapter 12.1-12.3]
• Trees and Forests (2/20/24)
  • Random Forests
  • [Murphy, Book 1, Chapter 18.1-18.5]
• Neural Networks (2/27/24)
  • Deep Feedforward Networks
• Latent Variable Models (3/5/24)
  • [Bishop PRML, Chapter 8.1-8.3]
• Computing with Latent Variable Models (3/12/24)
  • [Bishop PRML, Chapter 10.1-11.2]
• Deep Generative Models (3/26/24)
  • Automated Learning for Multivariate Diffusions
  • WaveNet
  • Stochastic Propagation and Approximate Inference in Deep Generative Models
• Causal Inference (4/2/24)
  • Statistics and Causal Inference
  • [Morgan and Winship, Counterfactuals and Causal Inference: Methods And Principles For Social Research (Analytical Methods for Social Research), 2nd ed, Chapter 2]
• Reinforcement Learning (4/9/24)
  • Sutton and Barto, Reinforcement Learning: An Introduction, Chapter 1 and Chapter 13
  • Human-level Control through Deep Reinforcement Learning
• Large Models in AI (4/16/24)
  • Language Models are Few-Shot Learners
  • Learning Transferable Visual Models from Natural Language Supervision
  • Fleuret, the Little Book of Deep Learning, Ch 3.2, 4.8, 4.9, 4.10 and 5.3
• Cautionary Tales in Machine Learning + Concept Reviews (Victory Lap)(4/23/24)
  • Hidden Stratification Causes Clinically Meaningful Failures in Machine Learning for Medical Imaging
• Class Presentations (4/30/24)

Assignments

In addition to the research project, a weekly reading response is due in class along with a couple of homework assignments. Students can summarize or focus and explain in detail some part(s) of the reading assignments (max one-page pdf). Students will be required to scribe one lecture in latex. This document should synthesize the key points in the lecture, where if another student reads it, they would understand the material from the class. A total of 7 late days can be used shared across all homework assignments.

• Homework 0 Open date: 1/30/24 Due date: 2/13/24

• Homework 1 Open date: 2/13/24 Due date: 2/27/24

• Homework 2 Open date: 3/5/24 Due date: 3/26/24

• Homework 3 Open date: 4/5/24 Due date: 4/25/24

Projects

There will be a research project due at the end of the term.

• Submit proposals 3/27/24

• Check-in with Staff: 4/1/24 - 4/10/24

• Project Presentations: 4/30/24

• Paper Submission: 5/9/24

Use of Generative AI

You can use Generative AI tools to get help with homework assignments and final project. However, you must report exactly in the submissions for which part of the assignment you used Generative AI and what initial prompt you used.

Office Hours

• Rajesh Ranganath: Tuesday 3:45-4:35pm, online
• Berfin Simsek: Monday 4:30-5:30pm, 60 5th Ave, #205C
• Xiang Pan: Wednesday 11:00am-1:00pm, 60 5th Ave, #737