Create an account to study the free first module. Start free.
Seven modules. Five synthesis checkpoints. One mock at the end. Each module has its own prereqs (skip if confident), core pages, and practice set — and references the material from earlier modules you’ll need to loop back to.
Linear algebra is the language of ML — every matrix equation is a sentence in shorthand.
Vector calculus turns those sentences into motions: gradient descent, backprop, optimization.
Probability and statistics layer on top to handle uncertainty. This course builds the first two.
Start with M1. Each module lists the prereqs you need (if any) and points back to earlier modules whose material it reuses.
Ax = b, Gaussian elimination to RREF, matrix inverses and the four equivalent invertibility conditions.
Span, basis, rank-nullity. Inner products, Gram-Schmidt, orthonormal basis.
Determinant & trace. Eigenvalues & eigenvectors deeper. Cholesky, diagonalization, SVD.
Gradients, Jacobian, matrix calculus, chain rule for vectors (backprop), Hessian & Taylor.
Unconstrained gradient descent, convexity, Lagrange multipliers, KKT.
SGD, learning rate schedules, momentum/AdaGrad/RMSProp/Adam, optimization landscape diagnosis.
PCA from two derivations. SVD↔PCA. SVM primal/dual, KKT, kernel trick.
Six pages that test combinations of two or three modules at once. Skip until you've finished the relevant cores.
For course navigation and interactive practice, please use the live app. Offline PDF generation is disabled.