A collection of machine learning experiments written entirely in modern C++—built from scratch for performance, transparency, and educational clarity. No frameworks. No dependencies. Just handcrafted logic that teaches how ML works under the hood.
| Module | Description |
|---|---|
xor_net.cpp |
Feedforward NN solving the XOR problem with backpropagation |
activations.hpp |
Sigmoid, ReLU, Tanh activation functions |
matrix.hpp |
Minimal matrix math utility for neural computation |
trainer.cpp |
Training loop with gradient descent and loss tracking |
Additional C++ modules are in progress (e.g. voice embedding, OpenCV classifiers, audio buffers)
Compile using any C++17 or newer compiler:
g++ xor_net.cpp -o xor_net -std=c++17
./xor_net