TITLE: Approximate Universal Artificial Intelligence

PRESENTER: Joel Veness, http://www.cse.unsw.com.au/~joelv/, joelv@cse.unsw.edu.au

AFFILIATION:School of Computer Science and Engineering, UNSW; NICTA, http://www.cse.unsw.edu.au

DATE: Friday 14th August 2009

TIME: 12:00:00

PLACE: CSE Seminar Room, Level 1, K17

ABSTRACT:

Marcus Hutte's AIXI agent provides a mathematically rigorous
optimality notion for general reinforcement learning agents. An
interesting open question is to what extent this ideal can be realised
in a computationally efficient manner. This talk focuses on some
promising (and soon to be published) recent work of the speaker that
has achieved impressive results on small, noisy, partially observable
and stochastic domains.

BIOGRAPHY OF SPEAKER:

Joel Veness is a 3rd year UNSW/NICTA funded PhD student. His interests
are Game Tree Search, Machine Learning for Games, Online Sequence
Prediction and Universal Artificial Intelligence. He enjoys
programming, and has previously written a general game playing
program, a computer chess program and a computer go program in the
name of entertainment.

Host:

Alan Blair

Seminar Convenor:

Van Hai Ho