Deep Learning - shivamvats/notes GitHub Wiki

Matrix Multiplication

Matrix-matrix multiplication is at the heart of both fully connected and convolutional neural networks. One efficient way to do is using GEMM (General Matrix to Matrix Multiplication). Convolution is infeasible without GEMM. For details, refer to https://petewarden.com/2015/04/20/why-gemm-is-at-the-heart-of-deep-learning/