search

toshikwa / gail-airl-ppo.pytorch

PyTorch implementation of GAIL and AIRL based on PPO.
MIT License
189 stars 30 forks source link

How to judge the convergence #3

Closed ghost closed 2 years ago

ghost commented 2 years ago

Hello!

Thanks so much for sharing the code!

I am new at inverse reinforcement learning. Now I am trying to apply AIRL and GAIL to a customized environment without knowing anything about the reward function. So are there any metrics that can be used to judge the convergence except for rewards?

Thanks ;).

toshikwa commented 2 years ago

Sorry for late response.

In general GAN training, it's hard to judge the convergence without watching the quality of Generator (e.g. generated images, return of the agent). If you want to judge the convergence based on the loss function (not return), you need to make the loss function meaningful. For example, you can read the paper about Wasserstein-GAN (WGAN).

Thanks.