Lecture Outlines and Supplementary Material, CO367/CM442 Winter'10

This webpage contains files used during the lectures. Additional material/links are also included. Be aware that this webpage is continually being polished/changed.

CO367/CM4420 is about Nonlinear Optimization and concentrates on Convex Optimization. In Winter'10 we are using the text Convex Optimization – Boyd and Vandenberghe. This book and the slides for the class lectures are available online. (This also includes videos of lectures given by Stephen Boyd in 2008. In addition, the old review sessions are useful.)
This course aims to cover parts (not all) of the eleven chapters from the text.
There will be six assignments that account for 40% of the final grade. These are due in class by 1:30PM on the due date. Late homework will not be accepted. You must work on your assignments on your own. Some of the assignments will use CVX. Please download and install this software.
There will be one midterm (20%) and a final exam (40%).

Lectures Date Subjects Covered Lecture and Supplementary Information
Lecture 35 M. Apr. 5 interior-point methods, pdf file; Ouline for final exam
Week 13 starts
Lecture 34 W. Mar. 31 interior-point methods, pdf file; barrier method (centering/duality gap);
generalized inequalities (SOCP, SDP); maxcut.m MATLAB file
Lecture 33 M. Mar. 29 interior-point methods, pdf file; Sections in text: 11.3, 11.4 Force field interpretation; barrier method (inner centering steps);
LP and GP examples; Feasibility/Phase I;
Week 12 starts
Lecture 32 F. Mar. 26 interior-point methods, pdf file; Sections in text: 11.3 p-d i-p method for linear programming (implementation details)
Lecture 31 W. Mar. 24 interior-point methods, pdf file; Sections in text: 11.3.2 p-d i-p method for linear programming (derivation)
Interpretation via perturbed KKT conditions

MATLAB program for p-d i-p for LP

Lecture 30 M. Mar. 22 interior-point methods, pdf file; Sections in text: 11.2.2, 11.3.1 More on log-barrier function (gradient/Hessian)
central path characterization/properties
primal/dual points on central path
Week 11 starts
Lecture 29 F. Mar. 19 interior-point methods, pdf file; Sections in text: 11.1, 11.2.1 Inequality constrained problems
optimality conditions (KKT conditions)
log-barrier function
Lecture 27 W. Mar. 17 Equality Constrained Minimization ( pdf file) Sections in text: 10.3.1, 10.3.2, 10.4.3 Newton step with elimination
infeasible start Newton's method
KKT for general equality constrained problem; and SQP methods with general equality constraints (nonconvex, not in text!!)
Lecture 26 M. Mar. 15 Equality Constrained Minimization ( pdf file) Sections in text: 10.1, 10.2 SQP approach with a merit function
Newton decrement
example: optimal allocation;
Newton step (equivalent to solution of quadratic program)
Newton decrement
Week 10 starts
Lecture 25 F. Mar. 12 Equality Constrained Minimization ( pdf file) Sections in text: 10. Solving KKT conditions;
quadratic programming
Lecture 24 W. Mar. 10 Equality Constrained Minimization ( pdf file) Sections in text: 10. problem 10.3 hint: use block elimination ;
linear and quadratic models for steepest descent and Newton directions;
affine invariance of Newton direction/step;
linear equality constraints; exploit optimality conditions; quadratic program (Hessian psd on nullspace of A!!!)
Lecture 23 M. Mar. 8 Equality Constrained Minimization ( pdf file) Sections in text: 10. review of Simple solution of Lagrangian relaxation equivalence of Boolean LP;
example of nonconvergence of steepest descent due to insufficient decrease.
Week 9 starts
Lecture 22 F. Mar. 5 Newton's method ( pdf file) Sections in text: 9.5 3 derivations (quadratic model, optimal scaling, linearization of optimality condition);
affine invariance; quadratic converence
Lecture 21 W. Mar. 3 steepest descent; sections in text: 9.4 intro, 9.4.1, 9.4.2, and only a very brief mention of the material in 9.4.3 and 9.4.4.
cancelled due to injury M. Mar. 1
Week 8 starts
Midterm F. Feb. 26 Lectures 1-18. Emphasis is on homework assignments and material covered in class.
Lecture 20 W. Feb. 24 Unconstrained Minimization ( pdf file) Sections in text: 9.? gradient descent (Cauchy's steepest descent); steepest descent on a quadratic function: steepdescquadr.m
Lecture 19 M. Feb. 22 Algorithms ( pdf file) Sections in text: 9.2 instability in solving linear equations: Aoneseps.m
Minimization problem solved with a log-barrier function: barrier.m
descent methods; line search; gradient descent (Cauchy's steepest descent); steepest descent on a quadratic function: steepdescquadr.m
Week 7 starts
Lecture 18 F. Feb. 12 Algorithms ( pdf file) Sections in text: 9.1 Unconstrained Minimization: iterative methods; closed sublevel sets; strong convexity
Lecture 17 W. Feb. 10 KKT Theorem and examples Finding the minimum eigenvalue and corresp. eigenvector (spectral theorem); finding a lower bound on the smallest eigenvalue
Lecture 16 M. Feb. 8 Duality ( pdf file) Sections in text: 5.3, 5.4, 5.5 TRS (trust region subproblem); geometric interpretation of duality; KKT conditions;
Remark on TRS
Week 6 starts
Lecture 15 F. Feb. 5 Duality Review of Homework #2
Lecture 14 W. Feb. 3 Duality CVX userguide, chapter 6, page 39, SDP.
matlab file for SDP relaxation of random max-cut problem
Lecture 13 M. Feb. 1 Duality ( pdf file) Sections in text: 5.2 weak duality p* ≥ d*; strong duality p* = d* AND d* is ATTAINED!; constraint qualifications guarantee a zero duality gap and dual attainment;
Week 5 starts
Lecture 12 F. Jan. 29 Duality ( pdf file) Sections in text: 5.1 Lagrangian (space where Lagrange multipliers lie); Lagrange dual function and lower bounds; Examples: linear least squares, LP, partitioning (max-cut), minimum volume covering ellipsoid.
Lecture 11 W. Jan. 27 Convex Programs ( pdf file) Sections in text: 4.6, 4.7 semidefinite programming; eigenvalue minimization; portfolio optimization (Pareto points)
Lecture 10 M. Jan. 25 Convex Programs ( pdf file) Sections in text: 4.4.2 (not minimal surfaces), 4.5.1-4.5.3 QCQP; SOCP; Robust programming
Geometric Progr.
Week 4 starts
Lecture 9 F. Jan. 22 Convex Programs ( pdf file) Sections in text: 4.4.1 (Voronoi diagrams from assignment, also for your interest only: finitely generated/polyhedral cones)
quadratic programming, least squares including: bounding variance, LP with random costs, Markowitz portfolio opt. (there are many sources for info. on portfolio opt., e.g. this with conic opt.)
Lecture 8 W. Jan. 20 Convex Optimization Problems ( pdf file) Sections in text: 4.2.-4.2.4 (read 4.2.5), 4.3 ( InfoSession on Grad Studies!)
Set midterm date for Feb 26, 2010.
download CVX
(gunzip and tar xvf OR unzip/xwinzip to get the cvx directory and follow the other installation instructions) create the startupcvx.m file for the appropriate path additions.
Try the cvx/matlab command quickstart.
Examples of Rockafellar-Pshenichni condition
Equivalent convex problems;
Lecture 7 M. Jan. 18 ( perspective, conjugate functions Sections in text 3.2.6, 3.3)
Convex Optimization Problems ( pdf file) Sections in text: 4.1, 4.2.2
( InfoSession on Grad Studies!)
    ( These old review sessions from EE364a 2009 may be useful. In particular: Review session 1 is on: examples of convex sets; affine/convex;conic hulls; preserving convexity; convex cones. Review session 2 is on: convex functions; determining convexity; preserving convex functions; conjugate functions. )
optimization problems in standard form
convex optimization problems
Rockafellar-Pshenichni optimality condition (nonnegative directional derivatives)
Week 3 starts
Lecture 6 F.Jan. 15 Convex functions cont... ( pdf file) Sections in text: 3.1.6 to 3.3.2, (and read 3.5 and 3.6) ( InfoSession on Grad Studies!)
further examples of characterizing convex functions;
epigraph and sublevel set; Jensen's inequality;
operations that preserve convexity;
(read text: perspective, conjugate functions, log-convex and log-concave functions, convexity with respect to generalized inequalities)
Lecture 5 W. Jan. 13 Convex functions ( pdf file); Sections in text: 3.1.1 to 3.1.5 Examples of convex functions and matlab file for plots;
Restriction to a line;
Characterizations of convex functions using: gradients, and using Hessians;
further examples
Lecture 4 M. Jan. 11 Dual Generalized Inequalities (pdf file) Dual Cone
Dual Generalized Inequalities
(dual minimum and minimal elements)
definition convex function;
Week 2 starts
Lecture 3 F. Jan. 8 (Duality) Convex sets cont... ( pdf file) separating (supporting) hyperplanes
Generalized Inequalities (minimum and minimal elements)
Lecture 2 W. Jan. 6 Convex Sets ( pdf file) Affine and convex sets; Convex Operations
Lecture 1 M. Jan. 4 Introduction to CO367/CM442 ( pdf file) Structure of Class
Math. Opt.; Least Squares; Nonliear Opt; Convex Opt;
CVX
Week 1 starts