We provide a tentative schedule. Please note that all course schedules are subject to change.
| Date | Lecture | Required Reading | Recommended Reading | Project |
| Apr 1 | Lecture 1, Intro + ML Knowledge Review - The Big Picture
- ML Concept Review
| |
Pattern Recognition and Machine Learning Ch 1-1.2.3 Deep Learning: Goodfellow Part II | |
| Apr 3 | Lecture 2 Biases and Fairness I + Project Overview - Sources of Biases
- Sensitive Features
- Fairness Through Unawareness / Awareness
- Statistical / Demographic Parity
| Barocas: Ch 2 | On formalizing fairness in prediction with machine learning: Gajane, 2017 | Project Guidelines Out |
| Apr 8 | Lecture 3 Biases and Fairness II - Conditional Statistical Parity
- Equality of Opportunity
- Counterfactual Fairness
- Preference-based Fairness
| | Equality of Opportunity in Supervised Learning: Hardt, 2016 | |
| Apr 10 | Lecture 4 Biases and Fairness III - Pre-processing Methods
- In-processing Methods
- Post-Processing Methods
- General ML Techniques
| | A survey on bias and fairness in machine learning: Mehrabi, 2019 | |
| Apr 15 | Lecture 5 Biases and Fairness IV - Fair Representation Learning
- Fair NLP
| | | |
| Apr 17 | Lecture 6 Biases and Fairness V - Fair Vision Learning
- Fairness in GDPR
| | The concept of fairness in the GDPR: a linguistic and contextual interpretation: Malgieri, 2020 | |
| Apr 22 | Lecture 7 Explainability, Interpretability and Transparency I - Intrinsic / Post Hoc Interpretability
- Model Specific / Model Agnostic Interpretability
- Local / Global Interpretability
- Evaluations
- Intrinsically Interpretable Models
| Molnar: Ch 2 | The mythos of model interpretability: Lipton, 2018 Explainable Machine Learning in Deployment: Bhatt et al. 2020 Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI: Arrieta, 2020 | |
| Apr 24 | Lecture 8 Explainability, Interpretability and Transparency II - Post Hoc : Model Agnostic Methods
- Explanation by Simplification
- Feature Interactions/Relevance
- ICE
- Local Surrogate (LIME)
- Sharply values
| Molnar: Ch 4 | | Project Proposal Due |
| Apr 29 | Lecture 9 Explainability, Interpretability and Transparency III - Post Hoc : Model Specific Methods
- Methods for Multi-Layered Perception Networks
- Methods for Convolutional Neural Networks
- Methods for Recurrent Neural Networks
| Molnar: Ch 5 | | |
| May 1 | Lecture 10 Explainability, Interpretability and Transparency IV - Example Based Methods
- Counterfactual Explanations
- Adversarial Examples
- Prototypes and Criticisms
- Influential Instances
| Molnar: Ch 6 | | |
| May 6 | Lecture 11 Explainability, Interpretability and Transparency V - Visualization Based Methods
- Network Visualization and Diagnosis
- Applications
| | | |
| May 8 | Lecture 12 Audits and Accountability I - Auditing AI Models
- AI principles
- Fairness Assessments
| | | |
| May 13 | Lecture 13 Audits and Accountability II - Model Risk Assessments
- NLP Biases Assessments
| | | |
| May 15 | Lecture 14 Robustness I - Adversarial Attacks
- Defense Methods
| Generative adversarial nets: Goodfellow 2014 | | |
| May 20 | Lecture 15 Robustness II - AI Deception
- Data Leakage Attack
| | | |
| May 22 | Lecture 16 Privacy - Differential Privacy
- Federated Learning
| | | |
| May 27 | Lecture 17 Applications - Recommendation Systems
- ML Concept Review
- Medical Diagnosis
- Hiring
- Search Engine
- Financial Lending
- Education
- Online Advertisement
- Human In the loop
| | | |
| May 29 | Lecture 18 Guest Lecture | | | |
| Jun 3 | Lecture 19 Final Project Presentation | | | Final Report Due
|