AIAI Seminar - 14 October 2024 - Dr Frank Wood

Title: Approximate Bayesian Inference, Algorithms, and Diffusion Models; Is Our Gig Up?

 

Abstract: I will start by sharing my take on the historical research arc of our community from hand-written graphical models and generic inference algorithms to model learning and amortized inference.  I will conclude by showing some inspiring (or worrying), recent results from using meta-learning and structured diffusion models for amortized approximate Bayesian inference.  Results that lead me to questions like "where's the model?" and "where's the inference algorithm?

 

Bio: Frank Wood is an associate professor of computer science at the University of British Columbia and a Canada CIFAR AI Chair at Mila. His primary research areas include deep generative modeling, amortized inference, probabilistic programming, reinforcement learning, and applied probabilistic machine learning. His research interests range from the development of new probabilistic models and inference algorithms to real-world applications, with contributions including probabilistic programming systems, new models and inference algorithms, and novel applications of such models to problems in autonomous driving, computational neuroscience, vision, natural language processing, robotics, and reinforcement learning. He directs the Programming Languages for Artificial Intelligence (PLAI) research group and is also a founder of Inverted AI, a PLAI group spin-out focused on advanced simulation technology for the autonomous vehicle industry.