1
$\begingroup$

Is there any place where people share their agent's settings for solving OpenAI Gym Environments?

For example, I'd like to know what are good parameters for a DDPG agent to learn the task in Reacher-v2. I believe that a lot of people tried to solve it and maybe they shared their solution for achieving better performance.

$\endgroup$

2 Answers 2

1
$\begingroup$

Specific hyperparameters for Reacher-v2, from Table 10 in Universal Successor Features for Transfer Reinforcement Learning:

Hyperparameter DDPG DDPG + USFs HER HER + USFs
Actor Learning Rate 1e-4 1e-4 1e-3 1e-4
Critic Learning Rate 1e-3 1e-3 1e-4 1e-3
Loss Weight λ N/A 1e-4 N/A 0.01
Batch Size 64 64 64 64
Discount Factor γ 0.99 0.99 0.99 0.99
HER Future Steps N/A N/A 50 50
HER Buffer Sampling Probability N/A N/A 0.5 0.5
$\endgroup$
0
$\begingroup$

This is kind of spread throughout the internet. There are tons of gists and repos, there are contests and topics on kaggle. I do not think there is one centralized repository of this sort of thing other than individuals who have curated some list(which is the repo I linked), although perhaps you could make one!

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .