Schedule

We will meet synchronously via Zoom: Tu 5:30-8:15

This course will proceed in three main parts: overview, deep dives, and wrap up.

Structure

Overview

In the first part of the course we will review ML basics, set norms for interaction and complete a survey of the topics that we will cover for the rest of the semester.

In this part of the class, Professor Brown will lead synchronous sessions. Students will be responsible for reading overviews, refreshing background material, and choosing an area for their course project. Students will start with an introductory demo or replication as a mini project.

Deep Dives

During the middle of the course we will spend one week on each topic. There will be 1-3 papers to read each week.

Students will be responsible for presenting papers in class on a rotating basis.

During this time students will have milestones where they need to complete interim steps for their course project. The first milestone will be a proposal that includes the specific products for the remainder of the milestones based on a template.

Conclusion

In the end of the course, we will focus on integrating ideas across multiple topics.

We will also workshop students’ projects, giving substantive feedback prior to the final submissions.

Final projects will be evaluated through a presentation and paper

Weekly topics

Table 1 Schedule

Class

Topic

Reading

Activities

2021-01-29

Introduction

None

introductions, expectation setting

2021-02-01

Probability Review

Model Based ML, chapter 1

reading discussion, setting

2021-02-03

ML Process & Mutual information preview

Scikit learn getting started,

live coding

2021-02-08

Missing Data: Intro strategies

Handling Missing Values when Applying Classification Models & Missing data imputation using statistical and machine learning methods in a real breast cancer problem

Paper discussion led by Daniel

2021-02-10

Missing data with graphical models and causal reasoning

Graphical Models for Inference with Missing Data & Missing Data as a Causal and Probabilistic Problem

Paper discussion led by Julian

2021-02-15

Current Challenges in Missing data

Handling Missing Data in Decision Trees: A Probabilistic Approach & How to miss data? Reinforcement learning for environments with high observation cost

Paper discussions by Xavier and Zhen

2021-02-17

Current Challenges in Missing data

How to deal with missing data in supervised deep learning

Paper discussion by Madhukara, Replication & testing discussion,

2021-02-22

Fairness

fairml classification chapter and friedler empricial comparison paper

Empirical setup

2021-02-24

Fairness

Reading

preview of lasso and admm constraint to multiobjecitve reformulation

2021-03-01

Multi-objective & constrained opt

Elastic Net

Paper presentation by Daniel, try out elastic net & LASSO in scikit learn

2021-03-03

Multi-objective & constrained opt

A critical review of multi-objective optimization in data mining: a position paper

Paper presentation and discussion by Zhen

2021-03-08

Latent Variable Models

Gaussian Mixture Models and Topic Models

Paper presentaiton by Xavier

2021-03-10

Latent Variable Models

Indian Buffet Process and Auto-Encoding Variational Bayes

Paper presentation by Madhukara

2021-03-15

Missing or Noisy labels

Learning with Noisy Labels and Semi Supervised Learning

Julian and Daniel

2021-03-17

Noisy Labels as a model for Bias

Recovering from biased data: Can fairness constraints improve accuracy and Fair classification with group dependent label noise

Zhen

2021-03-22

Interpretable & Explanation Intro

A Survey of Methods for Explaining Black Box Models

Xavier

2021-03-24

A Case for Interpretability over Explanation

Why are we explaining black box models and Learning Certifiably optimal rule lists for categorical data

Madhukara

2021-03-29

Models for Explanation

Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) and A unified approach to interpreting model predictions

Zhen

2021-03-31

Choosing Explanations and using explantions

How can I choose an explainer? An Application-grounded Evaluation of Post-hoc Explanations Actionable Recourse in Linear Classification

Daniel

2021-04-05

What are the risks of explanations

Model Reconstruction from Model Explanations

Xavier

2021-04-07

What does Interpretable mean

Towards A Rigorous Science of Interpretable Machine Learning and Towards falsifiable interpretability research

Madhukara

2021-04-12

Meta issues

The Scientific Method in the Science of Machine Learning and Value-laden Disciplinary Shifts in Machine Learning

Sarah

2021-04-13

Meta issues

Roles for computing in social change

Sarah

2021-04-19

Project Presentations

projects

presentations with peer feedback

2021-04-21

Project Presentations

projects

peer feedback

2021-04-26

Review and Project Reflections

Paper feedback

presentations with revision plans