Explore the development of deep learning models resistant to adversarial attacks in this 30-minute lecture from the University of Central Florida. Delve into key concepts such as security against attacks, minmax functions, and saddle point problems. Examine robustness strategies, analyze observations, and review experimental results to gain insights into creating more secure AI systems. Conclude with a summary of essential points for building resilient deep learning models in the face of potential adversarial threats.
Towards Deep Learning Models Resistant to Adversarial Attacks - CAP6412 Spring 2021
University of Central Florida via YouTube
Overview
Syllabus
Introduction
Security against attacks
Minmax function
Saddle point
Saddle point problem
Saddle point summary
Robustness
Observations
Experiments Results
Conclusion
Points
Taught by
UCF CRCV
Tags
Related Courses
-
Towards Deep Learning Models Resistant to Adversarial Attacks
-
Fast Is Better Than Free: Revisiting Adversarial Training - CAP6412 Spring 2021
-
Membership Inference Attacks against Adversarially Robust Deep Learning Models
-
Explaining and Harnessing Adversarial Examples in Machine Learning - Spring 2021
-
Cross-Domain Transferability of Adversarial Perturbations - CAP6412 Spring 2021
-
Defending Against Adversarial Model Attacks Using Kubeflow
