google / google/dopamine

About the dormant threshold setting in the ReDO paper

Open
#226 2 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
10.9k
Forks
1.4k
PR merge metrics
No merged PRs in 30d

Description

Hello there,

There seems to be an inconsistency on the default dormant threshold setting ($\tau$ in the paper) between the main paper and the appendix:

In the main paper (Page 5, Section 5), it says:
>For agents trained with ReDo, we use a threshold of τ = 0.1, unless otherwise noted, as we found this gave a better performance than using a threshold of 0 or 0.025.

In the appendix (Table 1), it says:
>0.025 for default setting, 0.1 otherwise

Could you please specify the best setting for this hyperparameter? Thanks!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.