EE 698W: Convex
Optimization in SP/COM
- Instructors:
Ketan Rajawat
- Units:
3-0-0-4
- Prerequisites:
EE605+EE621 OR Instructor Consent
- Objective:
Convex
optimization has recently been applied to a wide variety of problems in
EE,
especially in signal processing, communications, and networks. The aim
of this
course is to train the students in application and analysis of convex
optimization problems in signal processing and wireless communications.
At the
end of this course, the students are expected to:
- Know
about the applications of convex
optimization in signal processing, wireless communications, and
networking
research.
- Be
able to recognize convex optimization
problems arising in these areas.
- Be
able to recognize ‘hidden’ convexity
in many seemingly non-convex problems; formulate them as convex
problems.
- Be
able to develop low-complexity,
approximate solutions for difficult non-convex problems.
- References:
- Stephen Boyd and Lieven Vandenberghe, Convex
Optimization, Cambridge University Press. [Online]. http://www.stanford.edu/~boyd/cvxbook/
- IEEE
Signal Processing Magazine- Special Issue on Advances in Convex
Optimization, Vol. 27, No. 3, May 2010.
- Dimitri P. Bertsekas, Convex Analysis and Optimization,
Athena-Scientific, 2003.
- Format
(tentative)
- Major quiz (15) scheduled on 29 Jan
- Mid-sem exam (20) scheduled on Feb 22 in L16
- End-sem exam (30) scheduled on April 25 at 4-7pm in L16
- 2 Assignments (5x2) handed out on 15 Jan (due 22 Jan) and 3 March (due 12 March)
- 1 Computer Assignment (5x1) handed out on 30 Jan (due 26 Feb)
- 1 Term paper (20) handed out on 26 March (due 18 April)
- Time and place: MW 2-3:30pm, L12
- This course will cover (approximately):
- Background on linear algebra
- Convex sets, functions, and problems
- Examples of convex problems: LP, QCQP, SOCP
- Duality, KKT conditions
- Geometric programming and applications
- Linear and quadratic classification
- Network optimization
- Sparse regression, Lasso, ridge regression and applications in image processing
- Robust least squares and applications in signal processing
- Support vector machines and applications in machine learning
- Semidefinite programming and applications in experiment design
- Semidefinite relaxation and applications in MIMO detection, integer programming
- Low rank matrix completion and applications in recommendor systems
- Multidimensional scaling and applications in sensor localization
- Numerical linear algebra, basics of interior point methods
- Course material
- Proof that operator norm is a norm
- Assignment 1, Solutions to assignment 1
- Assignment 2, due 23 Jan
- Computer Assignment 1: 5.3, 5.4, 5.6, 5.7, 5.13 from Boyd's additional exercises due 26 Feb