Highly-scalable, physics-informed GANs for learning solutions of stochastic PDEs

10/29/2019
by   Liu Yang, et al.
42

Uncertainty quantification for forward and inverse problems is a central challenge across physical and biomedical disciplines. We address this challenge for the problem of modeling subsurface flow at the Hanford Site by combining stochastic computational models with observational data using physics-informed GAN models. The geographic extent, spatial heterogeneity, and multiple correlation length scales of the Hanford Site require training a computationally intensive GAN model to thousands of dimensions. We develop a hierarchical scheme for exploiting domain parallelism, map discriminators and generators to multiple GPUs, and employ efficient communication schemes to ensure training stability and convergence. We developed a highly optimized implementation of this scheme that scales to 27,500 NVIDIA Volta GPUs and 4584 nodes on the Summit supercomputer with a 93.1 peak and sustained half-precision rates of 1228 PF/s and 1207 PF/s.

READ FULL TEXT
research
02/19/2020

Physics-informed Neural Networks for Solving Nonlinear Diffusivity and Biot's equations

This paper presents the potential of applying physics-informed neural ne...
research
09/21/2018

Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems

Physics-informed neural networks (PINNs) have recently emerged as an alt...
research
06/06/2021

PID-GAN: A GAN Framework based on a Physics-informed Discriminator for Uncertainty Quantification with Physics

As applications of deep learning (DL) continue to seep into critical sci...
research
05/20/2020

Inverse Estimation of Elastic Modulus Using Physics-Informed Generative Adversarial Networks

While standard generative adversarial networks (GANs) rely solely on tra...
research
01/18/2023

Physics-informed Information Field Theory for Modeling Physical Systems with Uncertainty Quantification

Data-driven approaches coupled with physical knowledge are powerful tech...
research
10/19/2022

Bayesian Emulation for Computer Models with Multiple Partial Discontinuities

Computer models are widely used across a range of scientific disciplines...
research
10/12/2022

Overview on uncertainty quantification in traffic models via intrusive method

We consider traffic flow models at different scales of observation. Star...

Please sign up or login with your details

Forgot password? Click here to reset