https://github.com/andri27-ts/classiccartpole
CartPole using Policy Gradient Model Based
https://github.com/andri27-ts/classiccartpole
cartpole machine-learning neural-network policy-gradient python reinforcement-learning reinforcement-learning-algorithms
Last synced: 3 months ago
JSON representation
CartPole using Policy Gradient Model Based
- Host: GitHub
- URL: https://github.com/andri27-ts/classiccartpole
- Owner: andri27-ts
- Created: 2018-03-09T17:42:20.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2018-03-10T10:51:22.000Z (over 7 years ago)
- Last Synced: 2025-01-22T00:43:39.006Z (5 months ago)
- Topics: cartpole, machine-learning, neural-network, policy-gradient, python, reinforcement-learning, reinforcement-learning-algorithms
- Language: Python
- Homepage:
- Size: 5.86 KB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# ClassicCartPole
CartPole solved using policy gradient neural network trained on the model of the game.
### Motivation
This project is inspired by the amazing article about simple RL algorithms posted by Arthur Juliani https://medium.com/@awjuliani/simple-reinforcement-learning-with-tensorflow-part-3-model-based-rl-9a6fe0cce99
The basic ideas are the same as the post but implemented in a different way.
### How it works
In short, ClassicCartPole is constituted by two major parts: the model-based part and the policy-based part. The former try to learn the dynamics of the real environment, the latter the policy to take in every state.
The basic principle is that the policy is trained on the model instead of the real environment so that we actually never run the policy on the real environment.
In real tasks it's a huge opportunity to learn the model of the physical world, because you become capable to train the agent on the model of the environment, saving time and energy.
### Code structure
- [ModelNeuralNetwork.py](./ModelNeuralNetwork.py) : Contains the model neural network class
- [PolicyNeuralNetwork.py](./PolicyNeuralNetwork.py) : Contained the policy neural network class
- [PolicyModelBased.py](./PolicyModelBased.py) : Contains the main runner for the code. Here the model and the policy are trained and evaluated.
- [auxiliar.py](./auxiliar.py) : Contains some useful functions