Lec 02. How to Train a Neural Net
MIT OpenCourseWare · 79:34
This lecture explains how neural networks are actually trained: you search for parameters θ that minimize a cost (the sum of per-example losses) using first-order updates on a computational graph, with backpropagation...