Provably Efficient Lifelong Reinforcement Learning with Linear Function Approximation

06/01/2022
∙
by   Sanae Amani, et al.
∙
0
∙

We study lifelong reinforcement learning (RL) in a regret minimization setting of linear contextual Markov decision process (MDP), where the agent needs to learn a multi-task policy while solving a streaming sequence of tasks. We propose an algorithm, called UCB Lifelong Value Distillation (UCBlvd), that provably achieves sublinear regret for any sequence of tasks, which may be adaptively chosen based on the agent's past behaviors. Remarkably, our algorithm uses only sublinear number of planning calls, which means that the agent eventually learns a policy that is near optimal for multiple tasks (seen or unseen) without the need of deliberate planning. A key to this property is a new structural assumption that enables computation sharing across tasks during exploration. Specifically, for K task episodes of horizon H, our algorithm has a regret bound 𝒊Ėƒ(√((d^3+d^' d)H^4K)) based on 𝒊(dHlog(K)) number of planning calls, where d and d^' are the feature dimensions of the dynamics and rewards, respectively. This theoretical guarantee implies that our algorithm can enable a lifelong learning agent to accumulate experiences and learn to rapidly solve new tasks.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
∙ 10/15/2022

Near-Optimal Regret Bounds for Multi-batch Reinforcement Learning

In this paper, we study the episodic reinforcement learning (RL) probl...
research
∙ 12/30/2022

POMRL: No-Regret Learning-to-Plan with Increasing Horizons

We study the problem of planning under model uncertainty in an online me...
research
∙ 07/11/2023

Scaling Distributed Multi-task Reinforcement Learning with Experience Sharing

Recently, DARPA launched the ShELL program, which aims to explore how ex...
research
∙ 01/06/2023

Provable Reset-free Reinforcement Learning by No-Regret Reduction

Real-world reinforcement learning (RL) is often severely limited since t...
research
∙ 10/30/2019

Continuous Control with Contexts, Provably

A fundamental challenge in artificial intelligence is to build an agent ...
research
∙ 06/07/2017

Efficient Reinforcement Learning via Initial Pure Exploration

In several realistic situations, an interactive learning agent can pract...
research
∙ 06/24/2022

Joint Representation Training in Sequential Tasks with Shared Structure

Classical theory in reinforcement learning (RL) predominantly focuses on...

Please sign up or login with your details

Forgot password? Click here to reset