This project features a fully custom implementation of a Neural Network built entirely from scratch in C++, with no reliance on external machine learning libraries or frameworks. Its goal is to accurately classify handwritten digits from the well-known MNIST dataset.
- Pure C++ Implementation: No external ML libraries; all components (layers, neurons, activations, training, and inference) coded from scratch.
- Handwritten Digit Classification: Trained and evaluated on the MNIST dataset, achieving high accuracy.
- Interactive GUI Application: Includes an intuitive user interface allowing users to draw digits and receive instant predictions from the neural network.
This project demonstrates a deep understanding of neural network fundamentals, including forward propagation, backpropagation, activation functions, gradient descent, and network optimization techniques—all in raw C++.
2 Hidden 1 Output
Enjoy exploring and experimenting with this project!