# NASRL_DQN

**Repository Path**: frontxiang/NASRL_DQN

## Basic Information

- **Project Name**: NASRL_DQN
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2021-09-13
- **Last Updated**: 2021-09-13

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# NASRL_DQN
## NASRL_DQN is a PyTorch combination implementation of [Neural Architecture Search with Reinforcement Learning](https://arxiv.org/pdf/1611.01578.pdf) and [Deep Q learning](https://www.nature.com/articles/nature14236)

## Installation
```bash
git clone https://github.com/RL4RL/NASRL_DQN.git  
```
```python
pip install -r requirements.txt
```
Install [Pytorch](https://pytorch.org/)(v1.7)  
Add NASRL_DQN path into Python path: export PYTHONPATH=PATH_TO_NASRL_DQN:$PYTHONPATH  

## Running
### Train a single neural network on Atari datsets(Currently only for Pong game).
```python
python dql/dqn.py --conf block_conf.json
```

### Run Neural Architecture Search using RNN for multiple deep Q learning

```python
python nsarl/nas.py -c ./outputs/
```
### Run Neural Architecture Search using Evolutionary algorithm.
```python
python evo2.py
```
### Contributor (Alphabetical order)  
@[haotianteng](https://github.com/haotianteng)  
@[xcui297](https://github.com/xcui297)  
@[ypkuo708](https://github.com/ypkuo708)