Genetic Algorithm for Hyperparameter Optimization of a Convolutional Neural Network

This project implements a Genetic Algorithm (GA) to optimize the hyperparameters of a Convolutional Neural Network (CNN). The CNN is trained on the Sign Language MNIST (sign_mnist) dataset, which contains images of signed letters. The objective is to find the best combination of hyperparameters that maximizes the classification accuracy on the test dataset within a fixed number of epochs.

Features

• Hyperparameter Optimization: Utilizes a genetic algorithm to search for optimal hyperparameters.
• Deep Learning Framework: The CNN is built using PyTorch.
• Dataset: The sign_mnist dataset is used, consisting of hand-signed letters for image classification tasks.
• Performance-Oriented: The algorithm focuses on maximizing the test accuracy by fine-tuning critical hyperparameters.

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/genetic-algorithm-cnn.git
   cd genetic-algorithm-cnn```
   

2. Install the required dependencies:

   pip install -r requirements.txt```

Dataset

The sign_mnist dataset can be downloaded from Kaggle. Place the dataset in the appropriate directory (e.g., ./data) before running the training script.