Statistics 315a: Modern Statistical Learning

Lectures

Mondays and Wednesdays, 3:00PM – 4:30PM ECON 140

Contact and communication with staff

• Staff email list: stats315a-win2324-staff appropriate symbol lists.stanford.edu

• We will use Ed for discussion this quarter

• We will use Gradescope for grading and problem set submissions.

Instructor

John Duchi

• Office hours: Wednesdays, 4:30pm - 5:30pm, 126 Sequoia Hall.

Teaching Assistants

Zhaomeng Chen

• Office hours: Fridays, 10am - 12pm, 105 Sequoia Hall.

Chen Cheng

• Office hours: Thursdays, currently remote (Zoom link on Ed)

Zitong Yang

• Office hours: Tuesdays, 8am - 10am, 105 Sequoia Hall.

Prerequisites

Mathematical maturity and any convex combination of Stats 305A, 305B, CS229, or consent of instructor.

Description

In this course, we will provide an overview of modern techniques in supervised learning. This will include (among others)

• Regression methods, including linear regression and robust regression methods.

• Classification approaches, including linear discriminant analysis, logistic regression, and some perspective on neural networks

• Model selection and building methods, including boosting, stepwise methods, and basis expansions

• Kernel methods and reproducing kernel Hilbert spaces (RKHSs)

In addition to methods for actually making strong predictors, we will also discuss inferential tools, in this case meaning ways to guarantee validity of the different predictive models we have fit. This will include

• Parameter recovery and accuracy

• Predictive inference, including conformal methods and calibration

• Validation methods, including cross validation, bootsrap, and permutation testing.

Grading

Your grade will be determined by four problem sets (70%) and a final exam (30%). We reserve the right to change the grading rubric at any time.