improved noisy-networks

This is all the experimental code related to paper which aims to improve NoisyNet-DQN performance.

Environment

The code needs to run in the following environment：

python 3.5

pytorch 0.3.1

or

you can use the env.yml to generate the same virtual environment in anaconda:

conda create -f env.yml

conda activate subdeeprl

Script function

The script that trains a model using the model name as its file name.

The analysis script use analysis+params as its file name.

The result show script use output+ as its file name.

So you can see the function of the script from its name.