Computer Science 290/577

Seminar in AI: Optimization

Fall 2010

Instructor: Dan Gildea office hours MW 4-5pm
Time: Tu/Th 3:25-4:40, CSB 601
Motto: Meliora

Homework

Required text: Boyd and Vanderberghe, Convex Optimization

In addition to this text, we will cover additional numerical methods taken from Nocedal and Wright and Numerical Recipes. The second half of the course will focus on applications of optimization in the context of machine learning.

Th 9/2 IntroBoyd ch 1
Tu 9/7 Convex SetsBoyd ch 2
Th 9/9 Convex FunctionsBoyd ch 3
Tu 9/14 Convex Optimization ProblemsBoyd ch 4
Th 9/16 Convex Optimization Problems
Tu 9/21 DualityBoyd ch 5
Th 9/23 DualityBoyd ch 5
Tu 9/28 Approximation and Statistical EstimationBoyd ch 6, 7
Th 9/30 Geometric ProblemsBoyd ch 8
Tu 10/5 Unconstrained MinimizationBoyd ch 9
Th 10/7 Equality Constrained MinimizationBoyd ch 10
Tu 10/12Interior Point MethodsBoyd ch 11
Th 10/14Interior Point MethodsBoyd ch 11
Tu 10/19BFGSNocedal ch 8
Th 10/22L-BFGSNocedal ch 9
Tu 10/26Simulated AnnealingKirkpatrick
Th 10/28Deterministic AnnealingRose
Tu 11/2 Numerical MethodsPress 10.4 - 10.6
Th 11/4 Integer ProgrammingNemhauser ch II.4
Tu 11/9Machine Translation: MERTOch 2003, Cer
Th 11/11Convergence of EMXu and Jordan
Tu 11/16Image RestorationMersereau Lu
Th 11/18Dual Decomposition for ParsingRush et al. Koo et al.
Tu 11/23Dirichlet ProcessesGoldwater et al.
Th 11/25Thanksgiving break
Tu 11/30Exponentiated GradientBartlett
Th 12/2 Dirichlet Process IntroRanganathan
Tu 12/7 Variational Inference for Dirichlet Process MixturesBlei
Th 12/9 Project Presentations
Projects due: 5pm Friday 12/10

Grading

290:
  • Homework: 40%
  • Participation: 10%
  • Final project: 50%
577:
  • Homework: 30%
  • Lectures and participation: 30%
  • Final project: 40%

gildea @ cs rochester edu
December 2, 2010