This repository contains a genetic algorithm implementation for training a neural network, PaddleNet, to play a simple paddle game. The code is written in Python and uses PyTorch for the neural network and numpy for state representation.
- Python 3.x
- PyTorch
- numpy
-
Clone the repository:
git clone https://github.com/henilp105/GeneticPongAI.git cd GeneticPongAI -
Install the required packages:
pip install torch numpy
The game is a simple paddle game where a ball bounces around the screen, and the paddle must keep it from going off the screen. The paddle moves vertically, and the ball bounces off the walls and the paddle.
The PaddleNet class defines a neural network with three fully connected layers. The input to the network is the current state of the game, and the output is the movement of the paddle.
The GeneticAlgorithm class implements the genetic algorithm for training the PaddleNet. It includes methods for evaluation, selection, crossover, and mutation.
The training loop initializes the game and the genetic algorithm and runs for a specified number of generations. Each generation evaluates the population, selects the best individuals, and creates a new generation through crossover and mutation.