Homework 5 for Convex Optimization (PKU)
-
Solve by using CVX by calling different solvers mosek and gurobi.
-
Solve by calling mosek and gurobi directly.
(a) Projection gradient method by reformulating the primal problem as a quadratic program with box constraints
(b) Subgradient method for the primal problem
(c) Gradient method for the smoothed primal problem
(d) Fast gradient method for the smoothed primal problem
(e) Proximal gradient method for the primal problem
(f) Fast proximal gradient method for the primal problem
(g) Augmented Lagrangian method for the dual problem
(h) Alternating direction method of multipliers for the dual problem
(i) Alternating direction method of multipliers with linearization for the primal problem
- deterministic version of AdaGrad, Adam, RMSProp, Momentum