Lec 23. Metrized Deep Learning
MIT OpenCourseWare · 67:50
Neural-network theory should be assembled the same way networks are: characterize each layer’s sensitivity to input and weight perturbations, then compose those properties with explicit series/parallel rules so any ar...