Provably Efficient Imitation Learning from Observation Alone

TL;DR

This paper introduces FAIL, a new model-free algorithm for imitation learning from observations alone, which is provably efficient and effective in large-scale MDPs, extending the scope of sample-efficient learning.

Contribution

The paper presents the first provably efficient algorithm for IL from observations, capable of learning near-optimal policies with polynomial sample complexity independent of observation space size.

Findings

FAIL learns near-optimal policies in large-scale MDPs.

The algorithm's sample complexity is polynomial and independent of the number of observations.

FAIL demonstrates strong empirical performance on OpenAI Gym tasks.

Abstract

We study Imitation Learning (IL) from Observations alone (ILFO) in large-scale MDPs. While most IL algorithms rely on an expert to directly provide actions to the learner, in this setting the expert only supplies sequences of observations. We design a new model-free algorithm for ILFO, Forward Adversarial Imitation Learning (FAIL), which learns a sequence of time-dependent policies by minimizing an Integral Probability Metric between the observation distributions of the expert policy and the learner. FAIL is the first provably efficient algorithm in ILFO setting, which learns a near-optimal policy with a number of samples that is polynomial in all relevant parameters but independent of the number of unique observations. The resulting theory extends the domain of provably sample efficient learning algorithms beyond existing results, which typically only consider tabular reinforcement learning settings or settings that require access to a near-optimal reset distribution. We also investigate the extension of FAIL in a model-based setting. Finally we demonstrate the efficacy of FAIL on multiple OpenAI Gym control tasks.