Talk by Dr. Xingjun Ma (The University of Melbourne)

Speaker: Dr. Xingjun Ma

Title: On Adversarial Understanding, Detection and Defense

Abstract: Adversarial machine (deep) learning is an emerging research area which involves deep neural network (DNN) understanding, adversarial attack (crafting adversarial examples to fool DNNs), adversarial detection (detect whether or not a given test sample is adversarial), and adversarial defense (train DNNs that are inherently robust to adversarial examples). In this talk, I will introduce the background as well as some latest advances in adversarial research, based on two of our papers published at ICLR 2018 and ICML 2019. This will cover some geometric understanding of the adversarial subspace, and new insights on the min-max formulation of adversarial training.

Short bio: Dr. Xingjun Ma is a current research fellow, and a former PhD student at School of Computing and Information Systems, The University of Melbourne. He works on machine learning, adversarial deep learning and computer vision topics, and have published papers at top tier conferences such as ICML, ICLR, CVPR, IJCAI, AAAI, etc. Find out more about him: http://xingjunma.com/.