CNet: A Complex-Valued Deep Learning Framework with Wirtinger Autodifferentiation and FFT--Hadamard Convolution
CNet is a C++/CUDA framework for building deep complex-valued neural networks (CVNNs) using Wirtinger derivatives for optimization. It introduces a physics-native approach where classification uses Born rule measurements instead of softmax, enabling models that naturally handle intrinsically complex data like RF signals. The work shows this framework can match real-valued models on language tasks.