This project implements a simple neural network model using Python, with features that make it practical for real-world applications. The neural network is designed for educational purposes, allowing users to understand the basics of forward and backward propagation, model training, and prediction from scratch.
- Custom MLP Implementation: Build and train a Multi-Layer Perceptron (MLP) from scratch.
- Data Handling: Load data from CSV files, with automatic splitting into training and test sets.
- Training: Train the model using Mean Squared Error (MSE) loss with customizable learning rates and architecture.
- Prediction: Make predictions on new data after training the model.
- Model Evaluation: Evaluate the model using different metrics such as MSE, R² Score, and more.
- Experiment Tracking: Log training progress and save results for later analysis.
- Hyperparameter Tuning: Support for grid search and random search for hyperparameter tuning.
Make sure you have Python >= 3.8 installed on your machine.