Timezone: »

Learning to Score Behaviors for Guided Policy Optimization
Aldo Pacchiano · Jack Parker-Holder · Yunhao Tang · Krzysztof Choromanski · Anna Choromanska · Michael Jordan

Wed Jul 15 08:00 AM -- 08:45 AM & Wed Jul 15 07:00 PM -- 07:45 PM (PDT) @

We introduce a new approach for comparing reinforcement learning policies, using Wasserstein distances (WDs) in a newly defined latent behavioral space. We show that by utilizing the dual formulation of the WD, we can learn score functions over policy behaviors that can in turn be used to lead policy optimization towards (or away from) (un)desired behaviors. Combined with smoothed WDs, the dual formulation allows us to devise efficient algorithms that take stochastic gradient descent steps through WD regularizers. We incorporate these regularizers into two novel on-policy algorithms, Behavior-Guided Policy Gradient and Behavior-Guided Evolution Strategies, which we demonstrate can outperform existing methods in a variety of challenging environments. We also provide an open source demo.

Author Information

Aldo Pacchiano (UC Berkeley)
Jack Parker-Holder (University of Oxford)
Yunhao Tang (Columbia University)
Krzysztof Choromanski (Google)
Anna Choromanska (NYU Tandon School of Engineering)
Michael Jordan (UC Berkeley)

More from the Same Authors