CAP 6619 - U01
Advanced Topics in Machine Learning: Trust in AI
Instructor: Sumit Kumar Jha
Knights Foundation School of
Computing
College of Engineering and
Computing
Florida International University
Spring
Term 2024
Credit: 3
units
Meeting Information:
TuTh 12:30 PM - 1:45 PM, PG-6 144
Class
Number: 14896
University
Catalog Description:
Advanced course on machine learning principles and techniques. Students
propose, implement, and present a collaborative project with advanced machine
learning techniques. Prerequisite: CAP 5610.
Previous
edition with online video lectures and assignments for reference and historical
understanding of the material: https://sumitkumarjha.com/class2023/TrustinAI_HomePage.htm
Evaluation
|
Project
Proposal or Survey Proposal with Evaluation Metrics |
30% |
|
Project
or Survey Proposal and Current Status Presentation |
30% |
|
Student Paper
Presentation |
10% |
|
Student
Project Presentation with Evaluation Metrics |
29% |
|
Explainable
AI: Sci-Fi Example |
01% |
|
Optional
Assignments: 25% for each graded assignment |
125% |
|
Total |
100% |
Grades: A: 90%-100%, B: 75%-90%, C: 60%-75%,
D: 50%-60%, F: Below 50%.
Materials: Survey
Ideas and Templates, Project Ideas and
Templates
Tentative
Schedule
|
Week |
|
Topic |
Reading Material |
Optional Assignments |
|
1 |
January 9, 2024 January 11, 2024 |
1. Deep Inside Convolutional
Networks: Visualizing Image Classification Models and Saliency Maps 2. Grad-CAM: Visual Explanations from
Deep Networks via Gradient-based Localization |
Visualization Paper |
|
|
2 |
January 16, 2024 January 18, 2024 |
3. Axiomatic Attribution for Deep
Networks (Integrated Gradients) 4. SmoothGrad: Removing Noise by Adding Noise |
|
|
|
3 |
January 23, 2024 January 25, 2024 |
5. Fast Axiomatic Attribution for
Neural Networks 6. Integrated Decision Gradients:
Compute Your Attributions Where the Model Makes Its Decision. |
|
Optional Assignment 1 is Due |
|
4 |
January 30, 2024 February 1, 2024 |
7. Neural SDEs: Robust &
Explainable Analysis of Electromagnetic Unintended Radiated Emissions. 8. Self-Attention Attribution: Interpreting
Information Interactions Inside Transformer |
|
Optional Assignment 2 is Due |
|
5 |
February 6, 2024 February 8, 2024 |
|
|
Optional Assignment 3 is Due |
|
6 |
February 13, 2024 February 15, 2024 |
|
|
|
|
7 |
February 20, 2024 February 22, 2024 |
|
|
|
|
8 |
February 27, 2024 February 29, 2024 |
SPRING BREAK |
|
Optional Assignment 4 is Due |
|
9 |
March 5, 2024 March 7, 2024 |
9. IDGI:
A Framework to Eliminate Explanation Noise from Integrated Gradients |
|
Project/Survey Proposal is Due
(30% of class credit) Student Paper Presentation is Due
(10% of class credit) |
|
10 |
March 12, 2024 March 14, 2024 |
Student Proposal
Presentations Student
Proposal Presentations |
|
Optional Assignment 5 is Due |
|
11 |
March 19, 2024 March 21, 2024 |
Student
Paper Presentations |
|
|
|
12 |
March 26, 2024 March 28, 2024 |
Student
Paper Presentations |
|
Project or Survey Presentation is
Due (30% of class credit) |
|
13 |
April 2, 2024 April 4, 2024 |
11. Dehallucinating Large Language Models Using
Formal Methods Guided Iterative Prompting 12. Counterexample Guided Inductive
Synthesis Using Large Language Models and Satisfiability Solving |
|
|
|
14 |
April 9, 2024 April 11, 2024 |
13. Probabilistic Outputs for Support
Vector Machines and Comparisons to Regularized Likelihood Methods 14. On Calibration of Modern Neural
Networks Lecture on Shapley Values |
|
Student Project Presentation with
Evaluation Metrics is Due (29% of class credit) |
|
15 |
April 16, 2024 April 18, 2024 |
Student Project/Survey
Presentations 1. Vis 2. Ric 3. Tia 4. Raj 1. Joc 2. Bri 3. Ker 4. Nin 1. Fra 2. Sye 3. Nei 4. Moh |
|
|
|
16 |
April 23, 2024 |
Final Examination Week Class meeting
|
|
|
|
|
Additional Reading |
Attribution-Based Confidence Metric for Deep
Neural Networks |
|
