Dr. Abbas Khosravi, Associate ProfessorInstitute for Intelligent Systems Research and Innovation, Deakin University, Waurn Ponds Campus, Australia
Speech Title: Uncertainty quantification for artificial intelligence (UQ4AI): why and how
Abstract: As artificial intelligence (AI) technologies translate into real-world decision tools, many experts are questioning how much subject matter experts could trust decisions and predictions generated by these systems. Trust is the key mechanism that shapes how experts use and adopt AI. Currently, the lack of trust in AI systems is a significant drawback in the adoption of this technology in safety-critical and health-related applications. Confidence of an AI model, in particular deep learning models, about its output has been always a critical point to its performance and reliability. How can one develop, for example, a neural network that knows when it does not know? Answering this type of questions is a prerequisite for widespread deployment of neural networks in safety critical applications. The field of AI-based uncertainty quantification has gained an overwhelming attention among researchers in recent years resulting in an arsenal of different methods. Several methods and frameworks have been proposed in literature to generate predictive uncertainty estimates using neural networks.
In this workshop, we will provide a comprehensive overview on recent advances on deep learning-based uncertainty quantification techniques. This will include ensemble, Bayesian, and direct methods for deep uncertainty quantification. Applications of these algorithms for developing uncertainty-aware decision-making tools will be then reviewed and discussed.
Biography: Abbas Khosravi is an associate professor with the Institute for Intelligent Systems Research and Innovation, Deakin University, Australia. He completed his PhD in machine learning at Deakin University in 2010. His broad research interests include artificial intelligence, deep learning, and uncertainty quantification. He is currently researching and applying probabilistic deep learning ideas and uncertainty-aware solutions in healthcare and engineering domains. He has published more than 350 journal and conference papers and his h-index is 47 based on Google Scholar.