Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

The In-Sample Softmax for Offline Reinforcement Learning

About

Reinforcement learning (RL) agents can leverage batches of previously collected data to extract a reasonable control policy. An emerging issue in this offline RL setting, however, is that the bootstrapping update underlying many of our methods suffers from insufficient action-coverage: standard max operator may select a maximal action that has not been seen in the dataset. Bootstrapping from these inaccurate values can lead to overestimation and even divergence. There are a growing number of methods that attempt to approximate an \emph{in-sample} max, that only uses actions well-covered by the dataset. We highlight a simple fact: it is more straightforward to approximate an in-sample \emph{softmax} using only actions in the dataset. We show that policy iteration based on the in-sample softmax converges, and that for decreasing temperatures it approaches the in-sample max. We derive an In-Sample Actor-Critic (AC), using this in-sample softmax, and show that it is consistently better or comparable to existing offline RL methods, and is also well-suited to fine-tuning.

Chenjun Xiao, Han Wang, Yangchen Pan, Adam White, Martha White• 2023

Related benchmarks

TaskDatasetResultRank
Offline Reinforcement LearningD4RL (various)
HalfCheetah-Medium48.3
22
LocomotionMuJoCo walker2d medium-replay D4RL
Average Normalized Score69.8
16
LocomotionMuJoCo halfcheetah (medium)
Normalized Score48.3
10
LocomotionMuJoCo hopper (medium-replay)
Normalized Score92.1
10
LocomotionMuJoCo hopper medium
Normalized Score60.3
10
LocomotionMuJoCo walker2d medium
Normalized Score82.7
10
LocomotionMuJoCo halfcheetah (medium-replay)
Normalized Score44.3
10
LocomotionMuJoCo halfcheetah-medium-expert
Normalized Score83.5
5
LocomotionMuJoCo hopper-medium-expert
Normalized Score93.8
5
LocomotionMuJoCo walker2d-medium-expert
Normalized Score109
5
Showing 10 of 13 rows

Other info

Follow for update