Timezone: »

 
Poster
On the Value of Interaction and Function Approximation in Imitation Learning
Nived Rajaraman · Yanjun Han · Lin Yang · Jingbo Liu · Jiantao Jiao · Kannan Ramchandran

Wed Dec 08 04:30 PM -- 06:00 PM (PST) @ Virtual
in Poster Session 4 »
We study the statistical guarantees for the Imitation Learning (IL) problem in episodic MDPs.Rajaraman et al. (2020) show an information theoretic lower bound that in the worst case, a learner which can even actively query the expert policy suffers from a suboptimality growing quadratically in the length of the horizon, $H$. We study imitation learning under the $\mu$-recoverability assumption of Ross et al. (2011) which assumes that the difference in the $Q$-value under the expert policy across different actions in a state do not deviate beyond $\mu$ from the maximum. We show that the reduction proposed by Ross et al. (2010) is statistically optimal: the resulting algorithm upon interacting with the MDP for $N$ episodes results in a suboptimality bound of $\widetilde{\mathcal{O}} \left( \mu |\mathcal{S}| H / N \right)$ which we show is optimal up to log-factors. In contrast, we show that any algorithm which does not interact with the MDP and uses an offline dataset of $N$ expert trajectories must incur suboptimality growing as $\gtrsim |\mathcal{S}| H^2/N$ even under the $\mu$-recoverability assumption. This establishes a clear and provable separation of the minimax rates between the active setting and the no-interaction setting. We also study IL with linear function approximation. When the expert plays actions according to a linear classifier of known state-action features, we use the reduction to multi-class classification to show that with high probability, the suboptimality of behavior cloning is $\widetilde{O}(dH^2/N)$ given $N$ rollouts from the optimal policy. This is optimal up to log-factors but can be improved to $\widetilde{O}(dH/N)$ if we have a linear expert with parameter-sharing across time steps. In contrast, when the MDP transition structure is known to the learner such as in the case of simulators, we demonstrate fundamental differences compared to the tabular setting in terms of the performance of an optimal algorithm, Mimic-MD (Rajaraman et al. (2020)) when extended to the function approximation setting. Here, we introduce a new problem called confidence set linear classification, that can be used to construct sample-efficient IL algorithms.
Keywords: Reinforcement Learning and Planning · Machine Learning

Author Information

Nived Rajaraman (University of California, Berkeley)

Hi I am Nived, a 2nd year PhD student at the University of California, Berkeley. My research interests are broadly in high dimensional statistics. More recently I have been exploring the theory of reinforcement learning

Yanjun Han (Stanford University)
Lin Yang (UCLA)
Jingbo Liu (University of Illinois, Urbana Champaign)
Jiantao Jiao (University of California, Berkeley)
Kannan Ramchandran (UC Berkeley)

More from the Same Authors