Advanced Topics in Statistical Learning: Spring 2023

Stat 241B / CS 281B

Instructor: Ryan Tibshirani (ryantibs at berkeley dot edu)

GSI: Seunghoon Paik (shpaik at berkeley dot edu)

Class times: Tuesdays and Thursdays, 3:30-5pm, Tan 180

Office hours:
RT: Wednesdays, 3-4pm, Evans 417
SP: Thursdays, 5-6pm, Evans 444

Handy links:
• Syllabus
• GitHub repo (source files for lectures and homeworks)
• Ed discussion (for class discussions and announcements)
• bCourses (for grade-keeping and homework solutions)
Go to:   Schedule | Homework | Project | Supplementary notes | Other resources

Schedule

Here is the estimated class schedule. It is subject to change, depending on time and class interests.

Week 1: Jan 17 - Jan 19 Stat/ML in a nutshell (review) pdf, source
Week 2: Jan 24 - Jan 26 Nearest neighbors and kernels pdf, source
Week 3: Jan 31 - Feb 2 Splines and RKHS methods pdf, source
Week 4: Feb 7 - Feb 9 Minimax theory pdf, source Hw 1 due Fri Feb 10
Week 5: Feb 14 - Feb 16 Empirical process theory pdf, source
Week 6: Feb 21 - Feb 23 Lasso pdf, source
Week 7: Feb 28 - Mar 2 Ridge pdf, source Hw 2 due Fri Mar 3
Week 8: Mar 7 - Mar 9 Ridgeless pdf, source
Week 9: Mar 14 - Mar 16 Gradient flow pdf (outline)
Week 10: Mar 21 - Mar 23 Buffer/spillover Hw 3 due Fri Mar 24
Week 11: Mar 28 - Mar 30 (Spring break, no class)
Week 12: Apr 4 - Apr 6 Conformal prediction pdf, source
Week 13: Apr 11 - Apr 13 Conformal under distribution shift pdf, source Hw 3 due Fri Apr 14
Week 14: Apr 18 - Apr 20 Scoring and calibration pdf, source
Week 15: Apr 25 - Apr 27 Buffer/spillover
Week 16 May 2 - May 4 Class presentations Project due Fri May 5

Homework

Project

See here for instructions and timeline.

Supplementary notes

Here are some supplementary notes, on some topics adjacent to those from lectures.

Other resources

There is no course textbook. The lecture notes will be mostly self-contained, but will often provide references for further details on the topics they cover. Below are some general excellent references that may be helpful as well.