mathoptimizationfoundations
Convex optimization
Convex sets and functions, Lagrangian duality, KKT conditions — the theoretical foundation behind most ML losses.
Уровни глубины
L0Intro~4ч
Recognises convex vs non-convex problems; knows that convex problems have global minima.
L1Basics~20ч
Verifies convexity, solves small LP/QP problems, writes Lagrangian and checks KKT conditions.
L2Working~30ч
Derives dual problems; uses CVXPY/cvxopt to solve practical ML objectives (lasso, SVM).
L3Advanced~40ч
Analyses convergence rates of first- and second-order methods; proximal operators.
L4Research~80ч
Contributes new algorithms for structured non-convex or stochastic optimization.
Ресурсы
L1 — Basics
L2 — Working