Causal machine learning in a quantum world

Sally_Poster

Dr. Sally Shrapnel, University of Queensland
This colloquium will be held at 11 am, 30th August, in  Bldg.: Richards (05), Room #213
Abstract:
Despite impressive recent advances, current deep learning models have several shortcomings. They often learn spurious correlations, are vulnerable to adversarial attacks, are not robust to domain shifts, and fail to predict the effects of interventions. Many believe that some, if not all of these problems occur because such models fail to capture causal structure.
In this talk, I will explain what causal structure is, the methods we use to find it, and the ways in which these methods fail. On the way, I will highlight the fact that causal structure is usually regarded as a network of physical processes, responsible for generating meaningful correlations in data, and consistent with classical physics. However, we know classical physics to be a special case of quantum physics and until recently it was believed that a fully quantum account of causal structure was impossible.
In this talk, I will present a model of quantum causation that generalises ideas from conventional causal models. This work has foundational significance, gives a new perspective on quantum machine learning, and presents opportunities for future quantum technology. While I will cover some technical details of the framework, no prior knowledge of quantum theory is assumed.