This is a personal project for the initiation to real life applications of machine learning. Why not have some fun and create something meaningful?

├── data/                  # Dataset
│   └── breast-cancer.csv
├── src/                   # Core neural network code
│   ├── model.py           # Main entry point
│   ├── neuralnetwork.py   # Neural network layers and functions
│   ├── load_data.py       # Data loading utilities
│   ├── Saving_Weights.py  # Save weights to CSV
│   ├── loading_weights.py # Load weights from CSV
│   └── test.py            # Model evaluation
├── model_weights/         # Trained weights (CSV format)
├── notebooks/             # Jupyter notebooks for exploration
└── database/              # Optional MySQL storage (not required)

From the src/ directory:

cd src
python model.py

This will load the pre-trained weights from model_weights/ and evaluate the model on the test set, printing accuracy, precision, and recall.

In src/model.py, uncomment the training lines:

train(network, 1e-5, 2000, bce, bce_prime, X, Y, True)
save(network)

Then comment out load_weights(network) to avoid overwriting with old weights before training.

Weights are saved as CSV files in model_weights/:

• {layer_name}_weights.csv - weight matrices
• {layer_name}_bias.csv - bias vectors

This keeps things simple and human-readable. No need for pickle or database storage for a model this size.

• numpy
• pandas
• scikit-learn
• mysql-connector-python (only if using database features)
• I ran a local MySQL DB on my computer, so get your own server :|