General Atari 2600 Game Playing Michael Bowling Work with: Joel Veness, Marc Bellemare, Anna Koop, Mostafa Vafadoost http://www.arcadelearningenvironment.org Friday, September 14, 2012
Friday, September 14, 2012
http://www.arcadelearningenvironment.org Friday, September 14, 2012
Independent Many Varied Interesting Friday, September 14, 2012
Friday, September 14, 2012
Friday, September 14, 2012
Reinforcement Learning A.I. 0100010101...00001110101 Friday, September 14, 2012
Planning A.I. 0100010101...00001110101 Friday, September 14, 2012
Model Learning Model A.I. Friday, September 14, 2012
Imitation/Apprenticeship Learning Expert A.I. Friday, September 14, 2012
Transfer Learning A.I. Pitfall! . . . Pitfall II Friday, September 14, 2012
Intrinsic Motivation A.I. Friday, September 14, 2012
Friday, September 14, 2012
Training Testing Games Games Friday, September 14, 2012
Friday, September 14, 2012
(Bellemare et al., AAAI 2012) Contingency Awareness: knowing what you control Friday, September 14, 2012
(Bellemare et al., AAAI 2012) Contingency Awareness: knowing what you control Contingency Aware Unaware Friday, September 14, 2012
(Bellemare et al., AAAI 2012) Contingency Awareness: knowing what you control Inter-Algorithm Score Distribution 1.0 Fraction of Games Extended 0.5 Basic MaxCol Extended MaxCol 0.0 1.0 0.8 0.6 0.4 0.2 0 Inter-Algorithm Score Friday, September 14, 2012
(Bellemare et al., NIPS 2012) Sketch-Based Hashing: tug-of-war vs. standard hashing Hash Table Size: 1000 Hash Table Size: 5000 Hash Table Size: 20,000 1.0 1.0 1.0 Tug-of-War Tug-of-War Fraction of games Fraction of games Fraction of games Standard 0.5 0.5 0.5 Standard Tug-of-War Standard 0.0 0.0 0.0 1.0 0.8 0.6 0.4 0.2 0 1.0 0.8 0.6 0.4 0.2 0 1.0 0.8 0.6 0.4 0.2 0 Inter-algorithm score Inter-algorithm score Inter-algorithm score 55 Testing Games Friday, September 14, 2012
(Bellemare et al., In Prep) Model Learning: pixels, probabilities, and priors Friday, September 14, 2012
(Bellemare et al., In Prep) Model Learning: pixels, probabilities, and priors Friday, September 14, 2012
Questions? http://www.arcadelearningenvironment.org Source code for ALE and all agents available! Friday, September 14, 2012
Friday, September 14, 2012
Will there be a competition? No vs. Friday, September 14, 2012
Recommend
More recommend