Generalization Guarantees for Imitation Learning
arXiv:2008.01913
Abstract
Control policies from imitation learning can often fail to generalize to novel environments due to imperfect demonstrations or the inability of imitation learning algorithms to accurately infer the expert's policies. In this paper, we present rigorous generalization guarantees for imitation learning by leveraging the Probably Approximately Correct (PAC)-Bayes framework to provide upper bounds on the expected cost of policies in novel environments. We propose a two-stage training method where a latent policy distribution is first embedded with multi-modal expert behavior using a conditional variational autoencoder, and then "fine-tuned" in new training environments to explicitly optimize the generalization bound. We demonstrate strong generalization bounds and their tightness relative to empirical performance in simulation for (i) grasping diverse mugs, (ii) planar pushing with visual feedback, and (iii) vision-based indoor navigation, as well as through hardware experiments for the two manipulation tasks.
arXiv commentPresented at the Conference on Robot Learning (CoRL), 2020
References in corpus (11)
- Adam: A Method for Stochastic Optimization
- Semi-Supervised Learning with Deep Generative Models
- Evolution Strategies as a Scalable Alternative to Reinforcement Learning
- A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks
- Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data
- Reinforcement and Imitation Learning for Diverse Visuomotor Skills
- A Note on the PAC Bayesian Theorem
- Reinforcement Learning from Imperfect Demonstrations
- Stronger generalization bounds for deep nets via a compression approach
- Learning a Multi-Modal Policy via Imitating Demonstrations with Mixed Behaviors
- Probably Approximately Correct Vision-Based Planning using Motion Primitives