https://github.com/yaricom/rl-playground

The RL playground. Experiments with RL algorithms
https://github.com/yaricom/rl-playground

learning-agents openai reinforcement-learning rl-playground

Last synced: 11 days ago
JSON representation

The RL playground. Experiments with RL algorithms

• Host: GitHub
• URL: https://github.com/yaricom/rl-playground
• Owner: yaricom
• License: mit
• Created: 2016-05-09T21:00:38.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2016-05-10T14:16:09.000Z (over 8 years ago)
• Last Synced: 2024-11-05T15:51:49.357Z (about 2 months ago)
• Topics: learning-agents, openai, reinforcement-learning, rl-playground
• Language: Python
• Size: 7.81 KB
• Stars: 0
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# RL-playground

In this repository some of my experiments with Reinforcement Learning algorithms based on [OpenAi Gym ToolKit](https://gym.openai.com)

## Overview

Packages:

- [openai/envs](https://github.com/yaricom/RL-playground/tree/master/openai/envs) the OpenAi Gym compatible environments for evaluation

- [openai/agents](https://github.com/yaricom/RL-playground/tree/master/openai/agents) the learning agents

Environments:

- [NArmedBanditEnv](https://github.com/yaricom/RL-playground/blob/master/openai/envs/classic/narmedbandit.py) - N-armed bandit (stationary, nonstationary)

Learning agents:

- [SampleAverageActionValueAgent](https://github.com/yaricom/RL-playground/blob/master/openai/agents/sampleaverage.py) - the learning agent based on sample-average action-value selection algorithm for both stationary and nonstationary environments

## Usage

```Python

import gym

from openai.agents.sampleaverage import SampleAverageActionValueAgent

def main():

    # load environment

    env = gym.make('10ArmedBanditStationary-v0')

    # setup

    agent = SampleAverageActionValueAgent(num_actions = 10)

    episode_count = 1

    max_steps = 100

    reward = 0

    done = False

    for i in xrange(episode_count):

        ob = env.reset()

        for j in xrange(max_steps):

            action = agent.evaluate(reward, done)

            ob, reward, done, _ = env.step(action)

            if done:

                break

if __name__ == '__main__':

    main()

```