Syllabus

Unconstrained Optimization: line search method: Wolf condition, Goldstein condition, sufficient decrease and backtracking, Newtons method and Quazi Newton method; trust region method: the Cauchy point, algorithm based on Cauchy point, improving on the Cauchy point, the Dog- leg method, two-dimensional subspace reduction; nonlinear conjugate gradient method: the Fletcher Reeves method.

Constrained Optimization: penalty method, quadratic penalty method, convergence, non smooth penalty function, L1 penalty method, augmented Lagrangian method; quadratic programming, Schur complementary, null space method, active set method for convex QP; sequential quadratic programming, convex programming.