|
Tommi S. Jaakkola, Ph.D. Thomas Siebel Professor of Electrical Engineering and Computer Science and the Institute for Data, Systems, and Society MIT Computer Science and Artificial Intelligence Laboratory Stata Center, Bldg 32-G470 Cambridge, MA 02139 tommi at csail dot mit dot edu [home] [papers] [research] [courses] [people] |
This project based course introduces students to emerging opportunities of generative AI in scientific research with emphasis on methodology, rigorous evaluation, and scientific grounding. The course is designed to help formulate, develop, and assess effective uses of generative AI methods to advance scientific objectives across domains such as molecular engineering, biology, and related areas. The course draws on recent literature pertaining to representation and embedding of scientific knowledge, inverse design, closed-loop discovery, and agentic approaches to hypothesis generation and experimentation. Students propose new generative strategies and evaluate their limitations, reproducibility, and scientific value. Enrollment may be limited.
This project oriented course focuses on in-depth modeling of engineering tasks as machine learning problems. Emphasizes framing, method design, and interpretation of results. In comparison to the broader co-requisite 6.C01/6.C51, this 6-unit project oriented subject consists of deep dives into selected technical areas or engineering tasks involving supervised, exploratory, and generative uses of machine learning. Deep dives into technical areas such as missing data, fairness, interpretability, causal discovery; engineering tasks such as recommender systems, performance optimization, or automated design. This 6-unit subject must be taken together with the 6-unit core 6.C01/6.C51. Enrollment may be limited.
Principles, techniques, and algorithms in machine learning from the point of view of statistical inference; representation, generalization, and model selection; and methods such as linear/additive models, Bayesian methods, and neural networks. Recommended prerequisite: 6.3900 or other previous experience in machine learning. Enrollment may be limited.
The class focuses on modeling with machine learning methods with an eye towards applications in engineering and sciences. Introduction to modern machine learning methods, from supervised to unsupervised models, with an emphasis on newer neural approaches. Emphasis on the understanding of how and why the methods work from the point of view of modeling, and when they are applicable. Using concrete examples covers the formulation of machine learning tasks, adapting and extending methods to given problems, and how the methods can and should be evaluated. Students cannot receive credit without simultaneous completion of a 6- unit disciplinary module offered by various departments in the school of engineering.
