Skip to content

intsystems/2026-Project-185

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TEMPO: Regime-Aware Operator Learning with Locally Adapted POD Bases


License GitHub Contributors GitHub Issues GitHub Pull Requests

Author Rodion Akinzhala
Consultant Alexander Terentyev
Advisors Yury Maximov, DSc
Vadim Strijov, DSc

Assets

Abstract

Operator learning methods typically approximate the solution map of a physical system with a single global model — an assumption that fails when solutions change character fundamentally across the parameter space.

We propose TEMPO: a framework that identifies latent dynamical regimes in snapshot data via Expectation-Maximization, builds a dedicated Neural-POD basis for each, and learns to route new inputs to the right regime at inference time. On multi-regime benchmarks, TEMPO achieves substantially lower errors than POD-DeepONet and vanilla DeepONet, with interpretable structure that reflects the underlying physics.

Keywords

Operator learning, DeepONet, Proper Orthogonal Decomposition, Expectation-Maximization, Gaussian mixture models, Regime discovery

About

No description or website provided.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors