Timezone: »

Representations and Exploration for Deep Reinforcement Learning using Singular Value Decomposition
Yash Chandak · Shantanu Thakoor · Zhaohan Guo · Yunhao Tang · Remi Munos · Will Dabney · Diana Borsa

Thu Jul 27 01:30 PM -- 03:00 PM (PDT) @ Exhibit Hall 1 #110

Representation learning and exploration are among the key challenges for any deep reinforcement learning agent. In this work, we provide a singular value decomposition based method that can be used to obtain representations that preserve the underlying transition structure in the domain. Perhaps interestingly, we show that these representations also capture the relative frequency of state visitations, thereby providing an estimate for pseudo-counts for free. To scale this decomposition method to large-scale domains, we provide an algorithm that never requires building the transition matrix, can make use of deep networks, and also permits mini-batch training. Further, we draw inspiration from predictive state representations and extend our decomposition method to partially observable environments. With experiments on multi-task settings with partially observable domains, we show that the proposed method can not only learn useful representation on DM-Lab-30 environments (that have inputs involving language instructions, pixel images, rewards, among others) but it can also be effective at hard exploration tasks in DM-Hard-8 environments.

Author Information

Yash Chandak (Stanford University)
Shantanu Thakoor (DeepMind)
Zhaohan Guo (DeepMind)
Yunhao Tang (Google DeepMind)
Remi Munos (DeepMind)
Will Dabney (Google DeepMind)
Diana Borsa (Google DeepMind)

More from the Same Authors