CSE254WI11

From SeedWiki

CSE254 / Winter 2011

This course will be on the game theoretic approach to prediction. Topics covered will be Regret based learning, compression of individual sequences, the Hedge algorithms, learning in repeated games, the boosting game and Adaboost, boosting the margin, Uncountably infinite classes of experts, Blackwell's approachability theorem.

Evaluation

Each week there will be an assignment of 1-2 exercises from one of the books for the course. This will amount to 60% of the grade. The other 40% will be based on a final exam.

• Class Schedule
• Resources:
  • CSE291: seminar on online learning
  • Prediction, Learning and Games/ Cesa Bianchi and Lugosi
  • Boosting: Foundations and Algorithms, Schapire and Freund (book in preparation).
  • A probabilistic theory of Pattern Recognition/ Devroye, Gyorfi and Lugosi
• Discussion Board