Adversarial Approach to Prediction in Atari Games

Marian Andrecki, Nicholas Kenelm Taylor

Research output: Contribution to conferencePaper

26 Downloads (Pure)

Abstract

Recent advances in neural networks have resulted in reliable predictions of observations in Atari games. The quality of predictions can be improved by addition of a generative adversarial approach to the training. This research investigates this development and the benefits of unsupervised predictive learning for reinforcement learning agents.
Original languageEnglish
Number of pages2
Publication statusPublished - 7 Jun 2017
Event2017 EPSRC CDT Student Conference – Oxford, Bristol and Edinburgh - Oxford, United Kingdom
Duration: 7 Jun 20177 Jun 2017
http://aims.robots.ox.ac.uk/epsrc-cdt-student-conference-oxford-bristol-and-edinburgh/

Conference

Conference2017 EPSRC CDT Student Conference – Oxford, Bristol and Edinburgh
CountryUnited Kingdom
CityOxford
Period7/06/177/06/17
Internet address

Fingerprint Dive into the research topics of 'Adversarial Approach to Prediction in Atari Games'. Together they form a unique fingerprint.

  • Cite this

    Andrecki, M., & Taylor, N. K. (2017). Adversarial Approach to Prediction in Atari Games. Paper presented at 2017 EPSRC CDT Student Conference – Oxford, Bristol and Edinburgh, Oxford, United Kingdom.