C++ / Python · 2025
redCedar
A lightweight machine-learning framework built from scratch in C++, with Python bindings for constructing and training neural networks.
redCedar is an educational machine-learning framework built from first principles. Its C++ tensor implementation provides multidimensional storage, indexing, matrix multiplication, transposition, addition, reduction, and conversion to Python-native values.
The neural-network API includes linear layers, sigmoid and ReLU activations, mean squared error loss, and an optimizer. Pybind11 bindings expose the native implementation to Python, allowing models to be assembled and trained with a familiar high-level workflow.
The project is built with CMake and remains a work in progress, serving as a hands-on exploration of tensor systems, neural-network internals, computation graphs, and automatic differentiation.