Skip to content

Latest commit

 

History

History
17 lines (12 loc) · 950 Bytes

README.md

File metadata and controls

17 lines (12 loc) · 950 Bytes

Pendulum control using Reinforcement Learning

This project contains several scripts used to build, train and showcase a (deep) neural network on a Pendulum problem.

Models

All trained and untrained models are stored in the models directory by default. The name of the model is specified by the name of the subdirectory, which contains H5 files of untrained and trained model, and it's structure as an image.

Model Builder

Included is also an easy to modify model builder. Feel free to tweak network hyperparameters.

Agent

DQNAgent is a simple class that can load or save a specific model, decide upon the action, and remember and replay steps in the environment.

Pendulum Gym

The gym loads a specific model - either untrained or trained - as an agent and trains it on the pendulum environment.

Showcase

It is also possible to showcase a specific model in the pendulum environment without training it using a showcase script.