August 26, 2019: Seminar by Saumya Jetley, Oxford University, UK

DeepInsight - An examination of the class decision functions learned by deep nets
Saumya Jetley
Oxford University, UK

Date: Monday, August 26, 2019
Time: 3:30pm
Venue: ECSU Seminar Room, 9th Floor, S. N. Bose Bhavan


The talk will uncover some simplistic facets of the class score functions learned by deep image classification networks that explain their adversarial vulnerability and tie it to the celebrated performance of these nets. More particularly, our research identifies the fact that specific input image space directions tend to be associated with fixed class identities. This means that simply increasing the magnitude of correlation between any input image and a single image space direction causes the net to believe that more (or less) of the class is present. This provides a new perspective on the existence of universal adversarial perturbations for deep nets. Further, in the space mapped by the above directions, the adversarial vulnerability of the nets and their classification accuracy are closely entwined. Various notable and interesting observations emerge from this, with key implications for efforts to construct neural nets that are both accurate and robust to adversarial attack, and will be discussed in a greater detail during the talk.

Bio: Saumya Jetley has recently concluded her PhD degree under Prof. Philip Torr at the University of Oxford. Her research interests include machine vision, deep learning, attention models, adversarial vulnerability in deep classification networks, and more recently explainability and compressibility of deep nets at large. She joined the University of Oxford as a Masters (in research) student in 2014 and stayed on for her PhD. Previously, she has had 3 years of industrial research experience from working with a Govt. of India research organisation (C-DAC), was a summer intern at Xerox Research Centre Europe in 2015, and a visiting researcher at the autonomous driving start-up FiveAI in early 2019.

All are cordially invited.

Dipti Prasad Mukherjee

Head, Electronics and Communication Sciences Unit