EE5121: Convex Optimisation
January – May 2015
This is a post-graduate course, aimed at giving a rigorous introduction
to optimisation theory and algorithms.
Instructor: Krishna Jagannathan,
Assistant Professor
TAs: Sreelakshmi PM
<ee13s015@ee.iitm.ac.in>; Geetha C
<ee13s041@ee.iitm.ac.in>; Arjun Bhagoji
<arjunbhagoji@gmail.com>; Jainam Doshi <ee10b058@ee.iitm.ac.in>;
Slot: D
Location: ESB 350
Course Moodle
Pre-requisites: A good foundation in linear algebra and
real-analysis over Euclidean spaces
References:
1. Ben-Tal and Nemirovski, Lectures on Modern Convex Optimization,
available here.
2. Boyd and Vandenberghe, Convex Optimization, available here.
Course Contents:
Convex Analysis
o
Convex sets, cones, inner and outer descriptions
o
Topological properties
o
Polyhedral sets and representations
o
Some main theorems (Caratheodory, Krein-Milman...)
o
Separating and Supporting Hyperplanes
o
Convex functions, Jensen's inequality
o
Gradient inequality
o
Maxima and minima of convex functions
o
Sub-gradients
Linear Programming
o
Structural properties, theorem on alternative
o
LP Duality
o
Some applications in classification (machine learning), compressed
sensing, network flows...
The Art of Posing Convex Programs:
o
Conic programs, geometric programs (GPs), semi-definite programs (SDPs)
Convex Programming:
o
Convex theorem on alternative
o
Lagrangian, Duality
o
Optimality conditions (KKT Conditions, saddle points)
Basic Algorithms:
o
Gradient descent
o
Sub-gradient, conjugate-gradient
o
Newton's method, quasi-Newton methods