#量化神经网络

Theoretical framework explains why discontinuous STE training works for quantized neural networks, with practical implications for binary neural network implementations.