Quick jump links:    Announcements    Office Hours    Handouts etc.    Handouts (most recent class)    Homework Solutions

Jonathan Eckstein's Nonlinear Optimization Course 26:711:652

Announcements (As of December 17, 2019 08:01 AM)

• The course is now complete. I have e-mailed marked-up, graded versions of all submitted exams.  Course grades should be visible on the registrar site by December 18, 2019.
   
• Support and disability accommodation information is available.  Note that the "accommodation and support statement" in this document applies only to Rutgers-Newark students (I believe that a similar statement for New Brunswick students is under development). 
  

Usual Office Hour Schedule

My office is room 5145 of the business school on Livinston campus.  Office hours for this course are:

• Wednesday 2:00-4:00pm, September 11 - December 4
  • Office hours will be held Tuesday October 8, instead of Wednesday, October 9, due to Yom Kippur
  • Planned exceptions: On Thanksgiving week, office hours will be help Monday instead of Wednesday, and the class will be held Tuesday instead of Thursday
  • Planned cancelations: October 23 (planned travel)
• Additional times by appointment; contact me by e-mail for an appointment.

Check the announcements section above for office hour changes, extensions, additions, and cancelations.
 

Handouts, Class Materials, and Assignments

My presentation slides and logs of my tablet scribblings may be found on Blackboard.

1. Thursday, September 5.  Introduction, convex sets and functions, local and global optimality
   • Syllabus
   • Homework 1, due September 12
2. Thursday, September 12.  Differential behavior of convex functions, unconstrained optimality conditions, gradient-related algorithms
   • Notes on gradient-related algorithms are available from Blackboard (the treatment resembles that of the Bertsekas alternate book, so it is not public access)
   • No homework assignment this week
3. Thursday, September 19.  Sketch gradient-related convergence proof, overview of non-gradient unconstrained methods, convergence rate analysis, starting to use MATLAB for unconstrained algorithms, motivation for Newton methods
   • Introductory notes on the course MATLAB code base
   • First collection of MATLAB code (small updates made at 5pm, August 19, 2019)
   • Homework 2, due September 26
4. Thursday, September 26:  Newton methods, start conjugate gradient methods
   • Notes on theoretical convergence of Newton methods
   • Second collection of MATLAB code
   • Homework 3, due October 3
5. Thursday, October 3:  Conjugate gradient (started in previous class), spectral gradient, and pattern/compass search methods
   • Third MATLAB code collection is on BlackBoard (not on this site because some of the code is the same as the homework solution for this class)
6. Thurday, October 10:  Fundamentals of constrained optimality and constraint qualification
   • The first take-home exam is due in one week and available on Blackboard
   • funcdef10d.m file needed for the MATLAB part of the exam
7. Thursday, October 17:  Constraint qualifications, metric regularity, and Robinson's condition
8. Thursday, October 24:  Cones, Lagrange multipliers, and applications
   • Homework 5, due October 31
9. Thursday, October 31:  Lagrangian duality
10. Thursday, November 7:  Finish Lagrangian duality, start high-level overview of constrained optimization methods
    • Homework 6, due November 14
11. Thursday, November 14:  Newton barrier / interior point methods
    • Full code collection, including simple barrier method
12. Thursday, November 21:  Augmented Lagrangian methods, part 1
13. Tuesday, November 26:  Augmented Lagrangians part 2, operator splitting part 1
    • Full code collection, including a simple augmented Lagrangian method for equality constraints
14. Thursday, December 5:  Operator splitting part 2
    • The second take-home exam is available on Blackboard (modified at 2pm on December 5 to reflect postponed due date)
       

Homework/Exam Solutions

Homework and exam solutions will be posted on Blackboard, under password protection.