RL-for-TSP

Overview

This project compares solutions to the Traveling Salesman Problem (TSP) using both Q-Learning and a Policy Gradient approach with a neural network. The Q-Learning algorithm is based on tabular Q-values, while the Policy Gradient approach utilizes a neural network to learn a policy.

Table of Contents

• Getting Started
  • Prerequisites
  • Installation
• Usage
  • play-around
    • Q-Learning Parameters
    • Policy Gradient Parameters
• Results

Getting Started

Prerequisites

• Python (>=3.6)
• NumPy
• TensorFlow (for Q-Learning)
• PyTorch (for Policy Gradient)
• Matplotlib

Installation

Clone the repository:

 ```bash
 git clone https://github.com/your-username/traveling-salesman.git
 cd traveling-salesman

Usage

Run the file by using the folllowing command

```bash
python QL_PG.py

play-around

Change the parameters in the script and observe the graph.

Q-Learning Parameters

num_cities, num_episodes, epsilon, alpha, gamma 

Policay Gradient Parameters

num_cities, input_size, hidden_size, output_size, learning_rate, num_episodes

Results

View the results for loss over episdoes for Q-Learning and Policy Gradient Respectively.