22-Aug |
Slides
|
Course Intro and Logistics |
Ravi Iyer |
Presentation signup (via Campuswire) |
|
Reliability, Fairness and Ethics |
24-Aug |
Slides
|
Generative Pretrained Transformers (Large Language Models) Dependability Issues
|
Ravi Iyer |
|
Readings:
Improving Language Understanding by Generative Pre-Training
Challenges and Applications of Large Language Models
GPT-4 Technical Report
Sparks of Artificial General Intelligence: Early experiments with GPT-4
|
29-Aug |
Slides
|
Validation (Reliability Assessment) of Generative Language Models |
Ravi Iyer |
|
Background Readings: DECODINGTRUST: A Comprehensive Assessment of Trustworthiness in GPT Models
A Survey of Safety and Trustworthiness of Large Language Models through the Lens of Verification and Validation
|
31-Aug |
Slides
|
Trustworthy AI and Fairness |
Ravi Iyer, Student Presenters + Respondents |
|
Papers: Improving the Fairness of Chest X-ray Classifiers
Background Readings:
Fair ML Classification
On the Applicability of ML Fairness Notions
|
05-Sep |
Slides
|
Assessing bias as a measure of trustworthiness |
Ravi Iyer
|
|
Papers:
Towards Understanding and Mitigating Social Biases in Language Models
Readings: Improving fairness in machine learning systems: What do industry practitioners need?
The Landscape and Gaps in Open Source Fairness Toolkits
Finspector: A Human-Centered Visual Inspection Tool for Exploring and Comparing Biases among Foundation Models
|
07-Sep |
Slides
|
Reliability/Security: Game-theoretic models |
Ravi Iyer, Student presenter + respondent |
Project signup (via Campuswire) |
Paper:
Game-Theoretic Methods for Robustness, Security, and Resilience of CPS Control Systems
|
Robustness |
12-Sep |
Slides
|
Modelling uncertainty in AI/ML systems |
Ravi Iyer |
|
Paper:
Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation
Background Readings:
"Uncertainty in deep learning" Introduction: The Importance of Knowing What We Don’t Know
Bayesian Deep Learning
A Deeper Look into Aleatoric and Epistemic Uncertainty
Learning Dynamic Bayesian Networks
|
14-Sep |
|
Impact of distribution-shifts on learning in AI/ML systems |
Ravi Iyer, Student Presenter + Respondent |
|
Paper:
Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts?
Optional Readings:
Long tail challenge (Waymo)
On Pitfalls in OoD Detection: Predictive Entropy Considered Harmful
GEN: Pushing the Limits of Softmax-Based Out-of-Distribution Detection
Addressing AI tail cases |
19-Sep |
Recording |
Guest Lecture: Causal and Counterfactual Analysis for improving Robustness of DAG-based AI applications
|
Dr. Saurabh Jha (IBM Research) |
Lecture Critique |
|
21-Sep |
Slides |
Generative Adversarial Networks, Adversarial Robustness
|
Ravi Iyer (Intro), Student Presenter + Respondent |
|
Paper:
Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior
Optional Readings:
Generative adversarial nets (NIPS 2014)
PhysGAN: Generating Physical-World-Resilient Adversarial Examples for Autonomous Driving
DeepRoad: GAN-based Metamorphic Autonomous Driving System Testing
Generative Adversarial Networks for Black-Box API Attacks with Limited Training Data
Generative Adversarial Network for Wireless Signal Spoofing
Testing DNN-based Autonomous Driving Systems under Critical Environmental Conditions |
26-Sep |
Slides |
Group Activity |
Ravi Iyer + Student presenter |
|
|
28-Sep |
|
Group Activity - Solution Presentation |
Student Presenter
|
|
|
Verification and Certification |
03-Oct |
Slides
|
Fault-Injection |
Ravi Iyer (Intro), Student Presenter + Respondent
|
|
Paper ML-based Fault Injection for Autonomous Vehicles |
05-Oct |
|
Project proposal presentation |
Student presenter |
|
|
10-Oct |
Slides
|
Formal Verification Methods |
Ravi Iyer
|
|
Paper: Formal Scenario-Based Testing of Autonomous Vehicles: From Simulation to the Real World |
12-Oct |
|
Fuzz Testing |
Ravi Iyer |
|
Paper Reading: AV-FUZZER: Finding Safety Violations in Autonomous Driving Systems
Background Readings: Conformance Testing as Falsification for Cyber-Physical Systems |
Security/Privacy |
17-Oct |
|
AI-driven Malware Attacks |
Ravi Iyer(Intro), Student Presenter + Respondent |
|
Paper Ml-driven malware that targets av safety |
19-Oct |
|
Trojan Attacks and Stealing Models |
Student Presenter + Respondent |
|
Paper
PolicyCleanse: Backdoor Detection and Mitigation in Reinforcement Learning |
24-Oct |
|
Energy-based Models |
Ravi Iyer + Students |
|
Paper
A Tutorial on Energy-Based Learning (Slides)
A Tutorial on Energy-Based Learning (Report)
|
Explanability/ Interpretability |
26-Oct |
|
Counterfactual Reasoning: Introduction
Problem Definition and Application in Fault Sensitivity Assessment
Student Assessment of Assigned Papers
|
Ravi Iyer (Intro), Student Presenter + Respondent |
|
Paper
Exploring Counterfactual Explanations Through the Lens of Adversarial Examples: A Theoretical and Empirical Analysis
Optional Readings:
The Intriguing Relation Between Counterfactual Explanations and Adversarial Examples
Improving the accuracy of medical diagnosis with causal machine learning
If Only We Had Better Counterfactual Explanations: Five Key Deficits to Rectify in the Evaluation of Counterfactual XAI Techniques
|
31-Oct |
|
Guest Lecture: Adversarial Robustness and Certification |
|
Guest Lecture Critique |
Background Reading
|
02-Nov |
|
Interpretability |
Student Presenter + Respondent |
|
Paper: Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) Applications: Concept-based model explanations for Electronic Health Records
Human-Centered Tools for Coping with Imperfect Algorithms During Medical Decision-Making
Optional Reading: On the Interpretability of Artificial Intelligence in Radiology: Challenges and Opportunities
|
07-Nov |
|
Model Debugging |
Student Presenter + Respondent |
|
Paper Debugging Tests for Model Explanations
Optional Readings On Human Predictions with Explanations and Predictions of
Machine Learning Models: A Case Study on Deception Detection
|
09-Nov |
|
Midterm Project Presentation |
Student presenter |
Project Reports |
Details here |
New Problems |
14-Nov |
|
Causal Systems |
Ravi Iyer, Student Presenter + Respondent |
|
Paper: Sage: Practical & Scalable ML-Driven Performance Debugging in Microservices |
16-Nov |
|
Topic Review |
Ravi Iyer, Student Presenter + Respondent |
|
|
21-Nov |
|
FALL BREAK |
|
|
|
23-Nov |
|
FALL BREAK |
|
|
|
28-Nov |
|
Mechanistic Models |
Ravi Iyer |
|
Paper:
REMEDI: REinforcement learning-driven adaptive MEtabolism modeling of primary sclerosing cholangitis DIsease progression
Optional Reading
Reinforcement Learning based Disease Progression Model for Alzheimer's Disease
|
30-Nov |
|
Group Activity: Mechanistic Models + Bayesian Machine Learning |
|
|
|
05-Dec |
|
Evolving Themes: Impact of Generative Models in Healthcare Systems, Resilience and Trust
|
Ravi Iyer + Student presenter |
|
|
07-Dec |
|
Final Project Presentation |
Student Presenters |
|
|