Timezone: »

 
Poster
Instabilities of Offline RL with Pre-Trained Neural Representation
Ruosong Wang · Yifan Wu · Ruslan Salakhutdinov · Sham Kakade

Wed Jul 21 09:00 AM -- 11:00 AM (PDT) @

In offline reinforcement learning (RL), we seek to utilize offline data to evaluate (or learn) policies in scenarios where the data are collected from a distribution that substantially differs from that of the target policy to be evaluated. Recent theoretical advances have shown that such sample-efficient offline RL is indeed possible provided certain strong representational conditions hold, else there are lower bounds exhibiting exponential error amplification (in the problem horizon) unless the data collection distribution has only a mild distribution shift relative to the target policy. This work studies these issues from an empirical perspective to gauge how stable offline RL methods are. In particular, our methodology explores these ideas when using features from pre-trained neural networks, in the hope that these representations are powerful enough to permit sample efficient offline RL. Through extensive experiments on a range of tasks, we see that substantial error amplification does occur even when using such pre-trained representations (trained on the same task itself); we find offline RL is stable only under extremely mild distribution shift. The implications of these results, both from a theoretical and an empirical perspective, are that successful offline RL (where we seek to go beyond the low distribution shift regime) requires substantially stronger conditions beyond those which suffice for successful supervised learning.

Author Information

Ruosong Wang (Carnegie Mellon University)
Yifan Wu (Carnegie Mellon University)
Ruslan Salakhutdinov (Carnegie Mellen University)
Sham Kakade (University of Washington)

Sham Kakade is a Gordon McKay Professor of Computer Science and Statistics at Harvard University and a co-director of the recently announced Kempner Institute. He works on the mathematical foundations of machine learning and AI. Sham's thesis helped in laying the statistical foundations of reinforcement learning. With his collaborators, his additional contributions include: one of the first provably efficient policy search methods, Conservative Policy Iteration, for reinforcement learning; developing the mathematical foundations for the widely used linear bandit models and the Gaussian process bandit models; the tensor and spectral methodologies for provable estimation of latent variable models; the first sharp analysis of the perturbed gradient descent algorithm, along with the design and analysis of numerous other convex and non-convex algorithms. He is the recipient of the ICML Test of Time Award (2020), the IBM Pat Goldberg best paper award (in 2007), INFORMS Revenue Management and Pricing Prize (2014). He has been program chair for COLT 2011. Sham was an undergraduate at Caltech, where he studied physics and worked under the guidance of John Preskill in quantum computing. He then completed his Ph.D. in computational neuroscience at the Gatsby Unit at University College London, under the supervision of Peter Dayan. He was a postdoc at the Dept. of Computer Science, University of Pennsylvania , where he broadened his studies to include computational game theory and economics from the guidance of Michael Kearns. Sham has been a Principal Research Scientist at Microsoft Research, New England, an associate professor at the Department of Statistics, Wharton, UPenn, and an assistant professor at the Toyota Technological Institute at Chicago.

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors