Hybridised Temporal Difference Learning Ensemble with DQN Agents

This project implements a hybridised temporal difference learning ensemble using Deep Q-Network (DQN) agents. The ensemble approach combines model-free and model-based DQN agents to improve learning efficiency and performance in reinforcement learning tasks.

Requirements

• Python 3.11+
• Poetry for dependency management

Installation

1. Clone the repository:

   git clone https://github.com/H1drogen/Hybridised-TemporalDifferenceWeighted-Ensemble.git
   cd Hybridised-TemporalDifferenceWeighted-Ensemble

2. Install dependencies with Poetry:

   poetry install

Configuration

Edit the Environment settings and Hyperparameters in train_model.py

Edit Ensemble Hyperparameters in tdw/tdw_ensemble.py

Edit Agent Hyperparameters in DQN_Agent.py and DQN_Guided_Exploration.py

Usage

3. To train the agent, run:

   poetry run python train_model.py

4. To run Evaluation, run:

   poetry run python evaluate_model.py

calling any evaluation metrics with the right path to datasets.