An introduction to modern machine learning and data mining, grounded in the probabilistic methods behind contemporary AI.
Course Focus
Students study probabilistic modeling, uncertainty quantification, deep learning, generative models, and representation learning. The course connects mathematical and statistical principles to generative AI, multimodal learning, and foundation models. It also develops critical, effective use of AI tools in problem solving.
Assessment
- Midterm examination: 40%.
- Course project: 60%.
- Bonus homework: up to 20%.
The October 19 midterm starts with a two-hour closed-book written examination. A one-hour LLM-supported problem-solving portion follows, allowing students to revisit unsolved questions and tackle new ones. LLM use is required as part of the course, prohibited during the closed-book portion, permitted in the final hour, and must be disclosed whenever used.
Course Project
Teams of up to four use modern AI tools for a substantial technical project, such as investigating a research question or building a data-to-decisions pipeline. Evaluation emphasizes understanding the problem, identifying obstacles, and reaching concrete milestones rather than fully solving the question. The final written deliverable is a technical blog post, with project presentations at the end of the term.
Bonus homework offers challenging questions for students to explore with any suitable methods.
Tentative Weekly Outline
The first lecture is Tuesday, September 8, using the Monday class schedule. Subsequent meetings are on Mondays.
| Date | Topic |
|---|---|
| Tue, Sep 8 | Modern machine learning and probabilistic AI; Bayesian linear regression |
| Sep 14 | Gaussian processes |
| Sep 21 | Bayesian optimization |
| Sep 28 | Variational inference |
| Oct 5 | Bayesian deep learning |
| Oct 12 | Active learning and uncertainty quantification |
| Oct 19 | Midterm: two hours closed-book, followed by one hour of LLM-supported problem solving |
| Oct 26 | Hands-on tutorial: LLMs for problem solving |
| Nov 2 | Latent-variable generative models, VAEs, and diffusion |
| Nov 9 | Representation learning I: embeddings and autoencoders |
| Nov 16 | Representation learning II: self-supervised and predictive methods |
| Nov 23 | Contrastive learning, negative sampling, and current methods |
| Nov 30 | Multimodal representation learning and foundation models |
| Dec 7 | Project presentations and posters |