Show HN: CNET – Complex-valued neural networks from scratch in C++/CUDA

Category: ai-ml

Tags: complex-valued-neural-networks, cuda, c-plus-plus, wirtinger-derivatives, deep-learning-framework

Score: 6.3/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 3)

CNet is a C++/CUDA framework for complex-valued neural networks, featuring explicit Wirtinger gradients, spectral layers, and a Born-rule measurement head. It targets researchers exploring complex-valued learning and wave/field data, offering a transparent substrate with CPU and GPU implementations of every layer.

Target audience: AI researchers, computational physicists, signal processing engineers

Repository: https://github.com/crasmarum/CNet · C++ · MIT · 2 stars

View on Hacker News