Derivatives, integrals, gradients and the descent that uses them, ending at Newton's method.
9h · 11 lessons · 1 challenge
Before this: Linear Algebra for Machine Learning
Backpropagation is the chain rule applied to a graph, and nothing more. This module makes that literal: single-variable derivatives, then the chain rule, then integration, then the multivariable versions, then gradient descent, and only then the autograd engine that does it for you.
The second-order half comes last, as the source course orders it: curvature, the Hessian, convexity and Newton's method, which is what tells you why a loss curve goes flat.
Differentiate a two-layer network on paper, node by node, then check yourself against autograd. Scored on how far you get before you need a hint.
Worth 400 XP.
