cnn layers about rl_a3c_pytorch

[dgriff777]

As I see it as agent has to learn how to see and play the game I wanted to make sure agent had more then enough feature detections for images. The pooling was to reduce the size of overall model and is proven to be very effective with vision models. It wasn't as much cause v0 is harder but as I just thought the original cnn layers was insufficient and wanted to see how much it could learn if model had better tools to decipher scene.

The weights init are from the universe-starter-agent and borrowed pytorch implementation of it in pytorch-a3c repo. They seemed sufficient so saw no need to change.

[404akhan]

thanks for reply.

1. why not to use stride=2 to reduce size of model (repo pytorch-a3c did so), instead of pooling. max pooling lose spatial relation of where part of object came from, i think it is important in RL scene, though not yet tested influence.
2. i also saw at environment.py you use class NormalizedEnv, do you also think that using batchnorm layers will help after each cnn & possible after fc? i will try myself as well.

[dgriff777]

1. It's hard to really compare the use of CNN layers to pytorch-a3c as we don't have same input. Image is 80x80 whereas pytorch-a3c has 42x42. But I thought using stride 2 would be more ambiguous for model. I did try some configurations with stride 2 and didn't see it as superior.
2. I'm not sure I mean could speed things up but overall end performance wouldn't improve from doing that

[404akhan]

thx, please close an issue if you want