Zinovi Rabinovich, Assistant Professor, School of Computer Science – Carleton University<\/p>\n
Machine learning algorithms have been subjected to a range of attacks, both to thwart and to subvert their learning. It is particularly easy to do with Reinforcement Learning algorithms that heavily depend on their perceptions being reliable, their attempted actions correctly executed, and the rewards they reap indicative of the progress towards their goal. Control any one of those aspects, and you can make an RL agent fail or, worse, learn a bad behaviour. But what if perceptions come with error correcting codes, actions are verifiable, and the reward is strictly intrinsic to the agent? Are our RL agents safe from manipulation, then? Turns out no. It is possible, by the process of environment poisoning (i.e., changing how the environment behaves in response to agent actions), to manipulate an RL agent into learning a target (bad) behaviour. In this talk, I will show how it can be done, discuss how flexible the approach is, and what the future expects of it.<\/p>\n
Zinovi Rabinovich is an Assistant Professor in the School of Computer Science at Carleton University. He obtained his Ph.D. in Computer Science from the Hebrew University in Jerusalem, spent some years as an Algorithms Engineer at Mobileye Vision Technologies Ltd, and moved back into academia. His research focuses on how to leverage information asymmetry (in availability or in access) to manipulate decision processes. He’s looked into action advice provision, strategic information disclosure, election manipulation, and, more recently, poisoning Reinforcement Learning.<\/p>\n
https:\/\/scholar.google.com\/citations?user=JwJRnmAAAAAJ<\/a><\/p>\n https:\/\/dblp.org\/pid\/93\/4009.html<\/a><\/p>\n www.zinovi.net<\/a><\/p>\n <\/p>\n Seminar Moderator:<\/strong><\/p>\n