SP16: STOCHASTIC OPTMZTN FOR ML: 15811

Schedule

CS B659: Stochastic optimization for machine learning

Class schedule

This schedule is tentative and may change throughout the semester.

Additional readings/background can be obtained from Elad Hazan's free new textbook draft: (http://ocobook.cs.princeton.edu Links to an external site.)

A superset of the papers we will consider: Download readings.pdf

 

A useful reference for optimization: http://stanford.edu/~boyd/cvxbook/

 

Week Date Lecture and Readings Deadlines
1 Jan 11

Introductory lecture:  Download Lec1-Introduction.pdf

A nice set of slides relating different areas: http://ttic.uchicago.edu/~nati/Publications/ICML10tut.pdf

Background reading:

Chapters 1-3 of Hazan's book (http://ocobook.cs.princeton.edu)

"The Tradeoffs of Large Scale Learning", Bottou and Bousquet, 2008:  Download NIPS 2008 Bousquet.pdf

Additional (optional): "Online learning and online convex optimization", Shai Shalev-Shwartz, 2012:  Download FNT in Machine Learning 2012 Shalev-Shwartz.pdf

 

Assignment 1 released

2 Jan 18 No classes: MLK day
3 Jan 25

Discussion (with shared thought questions and discussion points from fellow classmates in Download readings_1.rtf

):

- "The Tradeoffs of Large Scale Learning", Bottou and Bousquet, 2008

- Online convex optimization framework

Readings for next week:

- "A stochastic gradient method with an exponential convergence rate for finite training sets", Roux et al., 2012:  Download NIPS 2012 Roux.pdf

- [Optional] 

"Non-strongly-convex smooth stochastic approximation with convergence rate O(1/n)", Bach and Moulines, 2013:  Download NIPS 2013 Bach.pdf

Assignment 1 due on Wednesday

Assignment 2 released

4 Feb 1

Discussion: (with shared thought questions and discussion points from fellow classmates in: readings_2.rtf Download readings_2.rtf)

"A stochastic gradient method with an exponential convergence rate for finite training sets", Roux et al., 2012:  Download NIPS 2012 Roux.pdf

Readings for next week:

"Taming the Monster: A Fast and Simple Algorithm for Contextual Bandits", Agarwal et al.  http://arxiv.org/abs/1402.0555

5 Feb 8

Discussion: (with shared thought questions and discussion points from fellow classmates in:  Download readings_3.rtf

)

"Taming the Monster: A Fast and Simple Algorithm for Contextual Bandits", Agarwal et al.  http://arxiv.org/abs/1402.0555 Links to an external site. Links to an external site.

Project discussions in pairs

Readings for next week:

 "Train faster, generalize better: Stability of stochastic gradient descent", Hardt et al., http://arxiv.org/pdf/1509.01240v1.pdf 

Assignment 2 due on Friday

6 Feb 15

Discussion (with shared thought questions and discussion points from fellow classmates in:  Download readings_4.rtf

): 

 "Train faster, generalize better: Stability of stochastic gradient descent", Hardt et al., http://arxiv.org/pdf/1509.01240v1.pdf

No new reading for next week

Talk assignment: Tor Lattimore's talk about online learning 

http://cs-colloq.soic.indiana.edu/future-talks/

Project Proposal due on Friday
7 Feb 22

Review previous readings from 2:30 - 3:30 p.m., with class moved to LH 325

Talk at 4 p.m.: Grigory Yaroslavtsev

http://cs-colloq.soic.indiana.edu/future-talks/

 

8 Feb 29

Student presentations about projects begin

Readings for next week:

 "Investigating practical, linear temporal difference learning", White and White, 2016:  Download white2016investigating.pdf

"Online dictionary learning for sparse coding", Mairal et al., 2009:  Download mairal2009online.pdf

 

Project presentation due
9 Mar 7

Discussion:

 "Investigating practical, linear temporal difference learning", White and White, 2016:  Download white2016investigating.pdf

"Online dictionary learning for sparse coding", Mairal et al., 2009:  Download mairal2009online.pdf

No new readings for after spring break

10 Mar 14 No classes: Spring break
11 Mar 21

Overview discussion topics:  Download discussion_march21.rtf

Readings for next week:

"Knows What It Knows: A Framework For Self-Aware Learning", Li et al., 2008:  Download Li08Knows.pdf

 

Assignment 2 results: Download results_assignment2.pdf

 

12 Mar 28

Discussion (with shared thought questions and discussion points from fellow classmates in:  Download readings_6.rtf

): 

"Knows What It Knows: A Framework For Self-Aware Learning", Li et al., 2008:  Download Li08Knows.pdf

Readings for next week:

"Fast Gradient-Descent Methods for Temporal-Difference Learning with Linear Function Approximation", Sutton et al., 2009

http://www.machinelearning.org/archive/icml2009/papers/546.pdf (Links to an external site.)

"Investigating practical, linear temporal difference learning", White and White, 2016:  Download white2016investigating.pdf
13 Apr 4

Discussion (with shared thought questions and discussion points from fellow classmates in: Download readings_7.rtf

): 

"Fast Gradient-Descent Methods for Temporal-Difference Learning with Linear Function Approximation", Sutton et al., 2009

http://www.machinelearning.org/archive/icml2009/papers/546.pdf (Links to an external site.)

"Investigating practical, linear temporal difference learning", White and White, 2016:  Download white2016investigating.pdf

Discuss projects for 1 hour (good initial draft for Monday)

14 Apr 11

Advice on projects:  Download Lec5-ProjectAdvice.pdf

Discussion on experimental design:  Download readings_8.rtf

Initial project draft due on Monday

Feedback on other projects due by end of the week

15 Apr 18

Project presentations (tentative ordering):

Vincent

Alex

Yangchen

Kai 

Erfan

Project presentations due
16 Apr 25

Project presentations (tentative ordering):

Bardia and Mohsen 

Yisu and Zhenhua

Eman and Srijita

Raksha and Tasneem

Project presentations due
17 May 2

 No Exam

Projects due