OLLIE: Imitation Learning from Offline Pretraining to Online Finetuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yue, Sheng, Hua, Xingyuan, Ren, Ju, Lin, Sen, Zhang, Junshan, Zhang, Yaoxue
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911893661155328
author Yue, Sheng
Hua, Xingyuan
Ren, Ju
Lin, Sen
Zhang, Junshan
Zhang, Yaoxue
author_facet Yue, Sheng
Hua, Xingyuan
Ren, Ju
Lin, Sen
Zhang, Junshan
Zhang, Yaoxue
contents In this paper, we study offline-to-online Imitation Learning (IL) that pretrains an imitation policy from static demonstration data, followed by fast finetuning with minimal environmental interaction. We find the naïve combination of existing offline IL and online IL methods tends to behave poorly in this context, because the initial discriminator (often used in online IL) operates randomly and discordantly against the policy initialization, leading to misguided policy optimization and $\textit{unlearning}$ of pretraining knowledge. To overcome this challenge, we propose a principled offline-to-online IL method, named $\texttt{OLLIE}$, that simultaneously learns a near-expert policy initialization along with an $\textit{aligned discriminator initialization}$, which can be seamlessly integrated into online IL, achieving smooth and fast finetuning. Empirically, $\texttt{OLLIE}$ consistently and significantly outperforms the baseline methods in $\textbf{20}$ challenging tasks, from continuous control to vision-based domains, in terms of performance, demonstration efficiency, and convergence speed. This work may serve as a foundation for further exploration of pretraining and finetuning in the context of IL.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17477
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle OLLIE: Imitation Learning from Offline Pretraining to Online Finetuning
Yue, Sheng
Hua, Xingyuan
Ren, Ju
Lin, Sen
Zhang, Junshan
Zhang, Yaoxue
Machine Learning
Artificial Intelligence
In this paper, we study offline-to-online Imitation Learning (IL) that pretrains an imitation policy from static demonstration data, followed by fast finetuning with minimal environmental interaction. We find the naïve combination of existing offline IL and online IL methods tends to behave poorly in this context, because the initial discriminator (often used in online IL) operates randomly and discordantly against the policy initialization, leading to misguided policy optimization and $\textit{unlearning}$ of pretraining knowledge. To overcome this challenge, we propose a principled offline-to-online IL method, named $\texttt{OLLIE}$, that simultaneously learns a near-expert policy initialization along with an $\textit{aligned discriminator initialization}$, which can be seamlessly integrated into online IL, achieving smooth and fast finetuning. Empirically, $\texttt{OLLIE}$ consistently and significantly outperforms the baseline methods in $\textbf{20}$ challenging tasks, from continuous control to vision-based domains, in terms of performance, demonstration efficiency, and convergence speed. This work may serve as a foundation for further exploration of pretraining and finetuning in the context of IL.
title OLLIE: Imitation Learning from Offline Pretraining to Online Finetuning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2405.17477