Faculty Candidate Seminar
Understanding How Deep Neural Networks See
Add to Google Calendar
CSE Lecturer Candidate Seminar
In this talk, we will introduce neural networks, and discuss why neural networks have been so successful in computer vision tasks. We will introduce the idea of Explainable AI, and discuss several ways of analyzing neural networks in order to understand how and why they produce the outputs that they do.
The talk is intended to be accessible to undergraduates with no machine learning background, although students a few weeks into a machine learning course will benefit more from the talk. The talk is adapted from lectures given in the third and fourth weeks of a third-year introductory course in neural networks and machine learning.
Michael Guerzhoy teaches in the Dept. of Computer Science at the University of Toronto and works in the Li Ka Shing Centre for Healthcare Analytics Research and Training (LKS-CHART) at St. Michael's Hospital in Toronto, where he is applying machine learning to patient data.
Michael has taught introductory courses in programming and CS theory and advanced courses in machine learning, computer vision, and data analysis at the University of Toronto. His curriculum development work includes co-designing the new Data Science undergraduate program at the University of Toronto, set to admit its first students in 2019, and co-designing the Computing for Medicine Initiative, a program for teaching computer science to students in the University of Toronto's School of Medicine, which has been running since 2016. Michael is particularly interested in assignment design for CS1/CS2 and for applied machine learning courses. Michael received the University of Toronto Computer Science Student Union (CSSU) Teaching Award in 2016 for his Introduction to Neural Networks course.
Michael takes on data science consulting projects, with clients ranging from the Canadian Broadcasting Corporation (CBC) to research groups at the University of Toronto. Michael has worked as an R&D engineer specializing in computer vision and machine learning at Epson in Toronto and at Inria in Grenoble, France.
In addition to applications of machine learning to healthcare data, Michael is interested in applied machine learning, computer vision, and applied statistics. Michael's work on modelling human travel (with Aaron Hertzmann) received the Best Paper Award at the Canadian Conference on Artificial Intelligence in 2014.
Michael holds a B.Sc. and M.Sc. in Computer Science from the University of Toronto, and an MSc in Statistics from the University of Toronto.