datvodinh / recurrent-ppo

A Reinforcement Learning Project using PPO + LSTM

PPO + LSTM Project

Introduction

This project explores the combination of Proximal Policy Optimization (PPO) and Long Short-Term Memory (LSTM) networks in reinforcement learning tasks. PPO is a popular policy optimization algorithm, while LSTM is a type of recurrent neural network that is capable of capturing temporal dependencies in sequential data. The goal of this project is to leverage the benefits of both PPO and LSTM to enhance the performance of reinforcement learning agents.

Graph

Forward Backward

Installation

1. Clone the repository:

git clone https://github.com/datvodinh10/recurrent-ppo.git

2. Install requirement:

pip install -r requirements.txt

Run

Open main.ipynb in Colab -> Run All

Update

• model_v2: split network, normalize state (running mean and var), use GRU (instead of LSTM)
• model: shared network, LSTM.