Type of Credit: Required
Credit(s)
Number of Students
This course will focus on the recent developments of Explainable and Safe AI. Explainable AI refers to a family of machine learning and analysis techniques that produce human-understandable machine learning models without sacrificing too much expressive power and prediction accuracy. For existing black box models, Explainable AI aims to generate post-hoc explanations capturing the cause and effect within the decision process of the model. Safe AI explores approaches to produce machine learning models that are robust to environmental noise and adversarial attacks. For trained models, Safe AI assesses if the model is stable and free of vulnerabilities. Explainable and Safe AI has been playing a critical role in realizing the ultimate goal of Trustworthy AI: artificial intelligence that is lawful, ethically adherent, and technically robust in each stage of its lifecycle, from design to development, deployment, and practice.
能力項目說明
The main course objective is to introduce up-to-date approaches for dealing with AI explainability and safety issues, equipping students with not only technical knowledge but also hands-on practical skills. At the end of this course, students should gain (1) general knowledge of the landscape of Explainable and Safe AI; (2) deep understanding of representative methods developed in the literature; (3) hands-on tool practice and development experience for enhancing machine learning explainability and safety.
| 教學週次Course Week | 彈性補充教學週次Flexible Supplemental Instruction Week | 彈性補充教學類別Flexible Supplemental Instruction Type |
|---|---|---|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Grading Scheme: Report 50% / Project 50%
Students are required to write a term report and finish a final project to demonstrate their understanding of the course materials.