Transfer learning is what happens when someone finds it much easier to learn to play chess having already learned to play checkers; or to recognize tables having already learned to recognize chairs; or to learn Spanish having already learned Italian. Achieving significant levels of transfer learning across tasks -- that is, achieving cumulative learning -- is perhaps the central problem facing machine learning.
This course will explore possible approaches to transfer learning, based on past work in machine learning (including computational learning theory) and statistics. Presentation will be shared among faculty, students, and some invited speakers. Students will carry out a substantial term project.
The course will serve to kick off a major DARPA project on transfer learning, involving faculty from Berkeley, Stanford, MIT, and Oregon State. It is expected that term projects may lead into thesis research areas supported by the project.
Reading: see the reading list for week-by-week topics and papers.