CO 769
Syllabus (Winter 2023)
Convex Relaxations of Numerically Hard Problems; Efficient Numerical Solutions
There will be four assignments with the last due on Fri. Apr. 14.
50% for the final take-home assignment/exam; and 50% for 3 assignments
during the semester.
The course will be self-contained.
The required theory will generally be covered when it arises.
We will cover (time permitting) elements from the following:
syllabus
-
Motivation
-
Role of cone optimization (LP/SDP/DNN) in solving hard numerical problems.
-
Examples of hard problems in discrete, combinatorial, engineering
optimization.
-
Background
- convex analysis
- semidefinite programming, SDP
- numerical linear algebra
- linear and
nonlinear programming optimality conditions
-
Interior point methods, implementations, applications
-
facial reduction and robustness
-
sparsity, chordal completions
-
First order methods, splitting methods
-
projections, fractional objectives
-
Applications and Implementations
-
low rank matrix completions,
Euclidean distance matrix completion
-
clustering, graph partitioning,
quadratic assignment, max-cut, quadratic knapsack,
-
molecular conformation, protein folding,
CO 769 webpage
Henry Wolkowicz, Department of Combinatorics and Optimization,
University of Waterloo, 200 University Ave. W., Waterloo, ON N2L 3G1.
(C) Copyright Henry Wolkowicz, 2022.
Last update: 01/11/2023 16:48:09, by Henry Wolkowicz