This project implements a simple neural network framework from scratch in Python, designed for educational purposes and experimentation. It does not rely on external deep learning libraries, providing a clear view of the underlying mechanics of neural networks.
main.py— Entry point for running and testing the neural network.src/activation.py— Contains activation functions (e.g., sigmoid, ReLU, etc.).src/layers.py— Implements neural network layers and their forward/backward passes.src/loss.py— Defines loss functions used for training.src/trainer.py— Handles the training loop, including forward, backward propagation, and parameter updates.requirements.txt— Lists required Python packages (minimal dependencies).
- Modular design for easy extension and experimentation.
- Customizable activation and loss functions.
- Manual implementation of forward and backward propagation.
- Simple training loop for supervised learning tasks.
- Install dependencies:
pip install -r requirements.txt
- Run the main script:
python main.py
This project is licensed under the MIT License.