"deep learning games"
Presenter: Dale Schuurmans, Professor, Department of Computing Science, University of Alberta; Principal Investigator, Amii
Abstract: "I will discuss a reduction of supervised learning to game playing that reveals new connections and provides some alternative training methods.
For convex one-layer problems, one can establish an equivalence between global minimizers of the training problem and Nash equilibria in a simple game. We have shown how the game can be extended to general acyclic neural networks with convex gates, and demonstrate a bijection between the Nash equilibria and critical (or KKT) points of the deep learning problem, assuming differentiability. Given this relationship we have been investigating alternative learning methods based on the no-regret algorithms used in computer game playing.
An interesting finding is that "regret matching'', a classical method from the game theory and economics literature, can achieve competitive training performance while producing sparser models than current deep learning algorithms."
Joint work with Martin Zinkevich
ai seminar series
Fridays at noon, Amii and the Department of Computing Science host AI Seminars, engaging presentations on topics in the broad field of artificial intelligence. With speakers from the University of Alberta and other world-leading groups, the talks give AI enthusiasts a friendly way of engaging with the latest trends and topics in research and development.
Seminars are open to the public, and no registration is required, though seating is limited and on a first-come-first-served basis. Topics range from foundational theoretical work to innovative applications of artificial intelligence technologies.
If you would like to present at an upcoming AI Seminar, please contact Colin Bellinger.
Join the AI Seminar mailing list to stay up-to-date on all the latest presentations