B9145: Reliable Statistical Learning

Hongseok Namkoong, Columbia University, Fall 2020

Description

As ML systems increasingly affect high-stakes decisions, it is critical that they maintain a reliable level of performance under operation. However, traditional modeling assumptions rarely hold in practice due to noisy inputs, shifts in environment, omitted variables, and even adversarial attacks. The standard machine learning paradigm that optimize average performance is brittle to even small distributional shifts, exhibiting poor performance on minority groups and tail inputs. Even performance of heavily engineered state-of-the-art models degrades significantly on domains that are slightly different from what the model was trained on. Lack of understanding of their failure modes highlights the need for models that reliably work, and rigorous safety tests to evaluate them.

This course surveys a range of emerging topics on reliability and robustness in machine learning. Most of the topics discussed in this class are active research areas, and relevant reading materials will draw upon recent literature (to be posted on the website). The goal of this class is to foster discussion on new research questions. This will encompass theoretical and methodological developments, modeling considerations, novel application areas, and other concerns rising out of practice.

Lectures

Tuesday, 3:50 - 7:05 PM in Uris 333 (see Canvas for Zoom link)

Course staff

Hongseok Namkoong (Instructor)

  • Email: namkoong (at) gsb.columbia.edu

  • Office hours: Wednesdays 4-5pm on Zoom

Chao Qin (TA)

  • Email: CQin22 (at) gsb.columbia.edu

  • Office hours: Fridays 4-5pm on Zoom

Prerequisites

There are no formal prerequisites, but the class will be fast-paced and will assume a strong background in machine learning, statistics, and optimization. This is a class intended for PhD students conducting research in related fields. Although some materials are of applied interest, this course has significant theoretical content that require mathematical maturity. The ability to read, write, and think rigorously is essential to understanding the material.

Grading

2 problem sets (50%) and a final project (50%).

Enrollment and HyFlex

Non-GSB students can follow this instruction to enroll in the class between Aug 28–Sep 4. Outside of this time window, email the instructor to enroll in the class.

This class will be in the HyFlex format. This means that the lectures will take place in Uris 333, and will be simultaneously broadcasted on Zoom (see Canvas for Zoom link). Enrolled students will have the option of physically attending lectures on certain dates. Email the course staff if you are enrolled in the class, but will only virtually attend lectures. Auditing students should join via Zoom.