We’re delighted to invite you to a seminar by Dr Matthew Wicker. Part of the UKRI Centre for Doctoral Training (CDT) in Safe and Trusted Artificial Intelligence (STAI) seminar series.
Abstract: The black-box nature of machine learning algorithms in conjunction with their many failure modes makes them difficult to practically deploy in safety-critical domains. Recent formal verification approaches address this by guaranteeing safe behavior. In this talk, we will focus on neural networks and will begin by demonstrating how we can formally rule out the existence of adversarial examples. We will then turn our attention to cutting-edge research that allows us to formally prove the robustness of a learning algorithm to adversarial attacks on its dataset.
Speaker Bio: Matthew Wicker is a Lecturer at Imperial College London and a Research Associate at the Alan Turing Institute. His group focuses on formal guarantees of trustworthiness for machine learning and artificially intelligent systems. Prior to his appointment at Imperial College, he received his PhD from the University of Oxford where he was a Google DeepMind scholar.