Explore the concept of adversarial training in machine learning through this 34-minute lecture from the University of Central Florida. Delve into the motivation behind revisiting adversarial training techniques, examining the technical contributions and background of the topic. Analyze the problems associated with Projected Gradient Descent (PGD) and compare various approaches. Review experimental results, including those on ImageNet, and consider the strengths and arguments presented in the conclusion. Gain valuable insights into the latest developments in adversarial machine learning and their implications for model robustness and security.
Fast Is Better Than Free: Revisiting Adversarial Training - CAP6412 Spring 2021
University of Central Florida via YouTube
Overview
Syllabus
Introduction
Motivation
Technical contribution
Background
Problem with PGD
Comparison
Experiments
Results
Imagenet
Conclusion
Strengths
Arguments
Taught by
UCF CRCV
Tags
Related Courses
-
Explaining and Harnessing Adversarial Examples in Machine Learning - Spring 2021
-
Robust Pre-Training by Adversarial Contrastive Learning - CAP6412 Spring 2021
-
Cross-Domain Transferability of Adversarial Perturbations - CAP6412 Spring 2021
-
Universal Adversarial Perturbations in Machine Learning - Spring 2021
-
Towards Deep Learning Models Resistant to Adversarial Attacks - CAP6412 Spring 2021
-
Trust Region Based Adversarial Attack on Neural Networks
