DERI Seminar with Albert Matveev, Senior Research Scientist at PhysicsX

Do join us on Thursday 29^th October at 11am with Albert Matveev, Senior Research Scientist at PhysicsX

Title: Light-Weight Diffusion Multiplier and Uncertainty Quantification for Fourier Neural Operators

Abstract: Operator learning is a powerful paradigm for solving partial differential equations, with Fourier Neural Operators (FNOs) serving as a widely adopted foundation. However, FNOs face significant scalability challenges due to overparametrization and offer no native uncertainty quantification (UQ), a key requirement for reliable scientific and engineering applications. Instead, neural operators rely on post hoc UQ methods that ignore geometric inductive biases. In this talk, I will present DINOZAUR, a diffusion-based neural operator parametrization with uncertainty quantification. Inspired by the structure of the heat kernel, DINOZAUR replaces the dense tensor multiplier in FNOs with a dimensionality-independent diffusion multiplier that has a single learnable time parameter per channel, drastically reducing parameter count and memory footprint without compromising predictive performance, while priors over these time parameters yield calibrated uncertainty estimates.

Bio: Albert Matveev is a Senior Research Scientist at PhysicsX, where he develops machine learning methods for modelling physics and geometry. His research spans 3D computer vision, geometric deep learning, and operator learning, with an emphasis on bringing theoretical advances into real-world engineering applications. He holds a PhD in Applied Mathematics from the Moscow Institute of Physics and Technology. Before joining PhysicsX, Albert worked on R&D projects at Huawei and Intel and conducted research at the Skolkovo Institute of Science and Technology. His work has been distinguished as a NeurIPS spotlight and as best dataset at the Eurographics Symposium on Geometry Processing.

________________________________________________________________________________

Microsoft Teams meeting

Join: https://teams.microsoft.com/meet/3642069876291696?p=X5a2LOwrvYpH6M0QZE

Meeting ID: 364 206 987 629 169 6

Passcode: rs6P93Qr

Need help? | System reference

For organizers: Meeting options

________________________________________________________________________________