MFNets: Learning network representations for multifidelity surrogate modeling

08/03/2020
by   Alex Gorodetsky, et al.
0

This paper presents an approach for constructing multifidelity surrogate models to simultaneously represent, and learn representations of, multiple information sources. The approach formulates a network of surrogate models whose relationships are defined via localized scalings and shifts. The network can have general structure, and can represent a significantly greater variety of modeling relationships than the hierarchical/recursive networks used in the current state of the art. We show empirically that this flexibility achieves greatest gains in the low-data regime, where the network structure must more efficiently leverage the connections between data sources to yield accurate predictions. We demonstrate our approach on four examples ranging from synthetic to physics-based simulation models. For the numerical test cases adopted here, we obtained an order-of-magnitude reduction in errors compared to multifidelity hierarchical and single-fidelity approaches.

READ FULL TEXT

page 1

page 14

page 15

page 17

page 18

research
06/22/2019

A support vector regression-based multi-fidelity surrogate model

Computational simulations with different fidelity have been widely used ...
research
12/10/2021

Surrogate-data-enriched Physics-Aware Neural Networks

Neural networks can be used as surrogates for PDE models. They can be ma...
research
11/16/2021

Adjoint-Matching Neural Network Surrogates for Fast 4D-Var Data Assimilation

The data assimilation procedures used in many operational numerical weat...
research
01/31/2023

A Data-Driven Modeling and Control Framework for Physics-Based Building Emulators

We present a data-driven modeling and control framework for physics-base...
research
09/22/2017

Hierarchical Kriging for multi-fidelity aero-servo-elastic simulators - Application to extreme loads on wind turbines

In the present work, we consider multi-fidelity surrogate modelling to f...
research
05/05/2023

Physics-Informed Localized Learning for Advection-Diffusion-Reaction Systems

The global push for new energy solutions, such as Geothermal, and Carbon...

Please sign up or login with your details

Forgot password? Click here to reset