STATS 608A: Optimization Methods in Statistics

Fall 2015

Class Information

Related Courses at Other Institutions

Instructor Information

GSI Information

Grading

Schedule

Project Teams

Class Information

Related Courses at Other Institutions

Instructor Information

Name: Ambuj Tewari

Office: 454 West Hall

Office Hours: By appointment

Email: tewaria@umich.edu

GSI Information

Name: Yun-Jhong Wu

Office Hours and Location: Mondays 7:30-8:30pm and Wednesdays 7:30-8:30pm in SLC (1720 Chemistry)

Email: yjwu@umich.edu

Grading

The final grade in the course will be determined by your scores in 3 homeworks and one final exam using the weights given below.

Schedule

Note: Schedule is subject to change. Links below will work if you have UM credentials.

Lecture number

Day

Topics

Remarks

1

Oct 26

Gradient Descent

2

Oct 28

Gradient Descent

Subgradients

3

Nov 2

Subgradients

Subgradient Method

4

Nov 4

Subgradient Method

Proximal Gradient Descent

HW 1 out

5

Nov 9

Proximal Gradient Descent

6

Nov 11

Newton’s Method

HW 1 due

HW 2 out

7

Nov 16

Coordinate Descent

Project proposals due

8

Nov 18

Conditional Gradient (Frank-Wolfe) Method

9

Nov 23

Conditional Gradient (Frank-Wolfe) Method

Proximal Newton Method

HW 2 due

Nov 25, 30

THANKSGIVING BREAK

HW 3 out, Nov 25

10

Dec 2

Fast Stochastic Methods

HW 3 due, Dec 5

11

Dec 7

Guest lecture: Online Convex Optimization

Guest lecturer: Sougata Chaudhuri

12

Dec 9

Guest lecture: Graphical Lasso

Guest lecturer: Kam Wong

13

Dec 14

Smoothing Methods

Dec 18

Project Reports Due

Project Teams

Team members

Topic

Robyn Ferg and Brook Luers

Inverse covariance matrix estimation

Jack Goetz (with Bopeng Li)

Clustering networks with node features

Pin-Yu Chen and Chun-Chen Tu

Feature selection for program execution optimization

Yura Kim and Wenyi Wu

Matrix decompositions using large-scale SDPs

Christopher Lee and Nick Seewald

Selection of optimal covariate values in experimental design

Joseph Bybee and Byoungwook Jang

Recurrent neural networks for text analysis

Joseph Dickens and Roger Fan

Deterministic annealing EM algorithm for mixture models

Xin Li and Ruofei “Brad” Zhao

Comparisons between traditional gradient methods and AdaGrad/AdaDelta

Julie Ghekas and Aritra Guha

Improvement of matching in observational studies using min-cost flow

Ziwei Cao and Timothy Lycurgus

Comparisons between online learning algorithms for regression