Uncertainty Quantification in Neural Network Models (25w5381)


Habib Najm (Sandia National Laboratories)

Prasanna Balaprakash (Oak Ridge National Laboratory)

Roger Ghanem (University of Southern California)

Serge Prudhomme (Polytechnique Montréal)

Yue Yu (Lehigh University)


The Banff International Research Station will host the “Uncertainty Quantification in Neural Network Models” workshop in Banff from February 16 - 21, 2025.

The spectacular advances in artificial intelligence (AI) and machine learning (ML) over the past decade have led to a continuing cascade of AI/ML developments and demonstrations. Moreover, this pace of development is nowhere near spent, with accelerating innovation leading to increasing AI/ML penetration in every aspect of society. These advances, however, are hardly ever disassociated from concerns about the degree of trust to be placed in predictions from AI/ML models. In large part, the strength of these models stems from their complexity and expressiveness. At the same time, this complexity can be a source of concern from the point of view of their reliability away from the training data. This concern is particularly relevant in applications in science, medicine, and engineering. It is clear that significant attention needs to be paid to assessing reliability of AI/ML models, and establishing estimates of uncertainty in their predictions, in order to provide confidence in their use in consequential applications where predictive errors can potentially lead to catastrophic results.

This workshop is focused on addressing the challenges inherent in the quantification of uncertainty in AI/ML models and their predictions, particularly in the context of neural network models. The meeting brings together experts from both the uncertainty quantification and machine learning communities, with the aim of enhancing communication and information exchange between the two disciplines, thereby fostering the development of novel ideas and formation of teams of researchers from diverse backgrounds to drive advances in this broad and critical area.

The Banff International Research Station for Mathematical Innovation and Discovery (BIRS) is a collaborative Canada-US-Mexico venture that provides an environment for creative interaction as well as the exchange of ideas, knowledge, and methods within the Mathematical Sciences, with related disciplines and with industry. The research station is located at The Banff Centre in Alberta and is supported by Canada’s Natural Science and Engineering Research Council (NSERC), the U.S. National Science Foundation (NSF), and Alberta’s Advanced Education and Technology.