FrozenLake-Q-Learning-Project

demonstrates how to use Q-Learning to solve the FrozenLake environment from OpenAI Gym. It includes advanced features such as logging, configuration files, model saving/loading, plotting, and hyperparameter tuning.

Requirements

• Python 3.6+
• numpy
• gym
• matplotlib
• scikit-optimize
• pyyaml

Project Structure

• configs/: Contains configuration files.
• logs/: Contains logs and saved data.
• models/: Contains saved models.
• plots/: Contains generated plots.
• src/: Contains source code.

Setup

1. Install the required packages:

   pip install -r requirements.txt
   

Instructions

Configure the Hyperparameters

Configure the hyperparameters in configs/params.json.

Run the Training Script

```bash
python src/main.py --mode train

Run the Evaluation Script

```bash
python src/main.py --mode evaluate

Configuration

• configs/config.json: Contains environment and file paths configurations.
• configs/params.json: Contains hyperparameters for the Q-Learning algorithm.

Logging and Plotting

• Logs are saved in the logs/ directory.
• Training progress plots are saved in the plots/ directory.

Advanced Features

• Logging: Logs training progress to a file.
• Configuration Files: Use JSON for environment and hyperparameter configurations.
• Model Saving and Loading: Save and load the Q-table to and from a file.
• Plotting: Generate plots for training rewards and episode lengths.
• Hyperparameter Tuning: Supports basic hyperparameter tuning using a configuration file.

Example Usage

Training the Agent

To train the Q-Learning agent, run:

```bash
python src/main.py --mode train

Evaluating the Agent

To evaluate the trained Q-Learning agent, run:

```bash
python src/main.py --mode evaluate