Posted: 28th February 2017

Shimon Whiteson has been announced as a recipient of a Google Research Award. Google puts out an annual open call for proposals on computer science and related topics including machine learning, machine perception, natural language processing, and security. The grants cover tuition for a graduate student and provide both faculty and students the opportunity to work directly with Google researchers and engineers.

For this round of funding Google received 876 proposals covering 44 countries and over 300 universities. Shimon’s proposal on, "End to End Deep Model Based Reinforcement Learning’ was one of only 143 successful proposals. 

Shimon explains, ‘Deep reinforcement learning has had some exciting successes in the past few years, but so far all those successes have been “model-free” i.e. they work without building an explicit model of the world, even though such a model has large potential advantages.  We have a hypothesis as to why model-based approaches to deep reinforcement learning are failing, and in this research we will build a new method that, if our hypothesis is correct, will enable model-based approaches to deep reinforcement learning to succeed.'