Graph-Informed Neural Networks

Søren Taverniers; Eric Joseph Hall; Markos A. Katsoulakis; Daniel M. Tartakovsky

Graph-Informed Neural Networks

Søren Taverniers
, Eric Joseph Hall
, Markos A. Katsoulakis
, Daniel M. Tartakovsky

Research output: Contribution to conference › Paper › peer-review

252 Downloads (Pure)

Abstract

Graph-Informed Neural Networks (GINNs) present a strategy for incorporating domain knowledge into scientific machine learning for complex physical systems. The construction utilizes probabilistic graphical models (PGMs) to incorporate expert knowledge, available data, constraints, etc. with physics-based models such as systems of ordinary and partial differential equations (ODEs and PDEs). Computationally intensive nodes in this hybrid model are replaced by the hidden nodes of a neural network (i.e., learned features). Once trained, the resulting GINN surrogate can cheaply generate physically-relevant predictions at scale thereby enabling robust sensitivity analysis and uncertainty quantification (UQ). As proof of concept, we build a GINN for a multiscale model of electrical double-layer capacitor dynamics embedded into a Bayesian network (BN) PDE hybrid model.

Original language	English
Number of pages	4
Publication status	Published - Mar 2021
Event	AAAI 2021 Spring Symposium Series: Combining Artificial Intelligence and Machine Learning with Physics Sciences - Stanford University (virtual), Palo Alto, United States Duration: 22 Mar 2021 → 24 Mar 2021 https://sites.google.com/view/aaai-mlps

Conference

Conference	AAAI 2021 Spring Symposium Series
Abbreviated title	AAAI-MLPS 2021
Country/Territory	United States
City	Palo Alto
Period	22/03/21 → 24/03/21
Internet address	https://sites.google.com/view/aaai-mlps

ASJC Scopus subject areas

General Computer Science

Access to Document

Final published versionFinal published version, 1.03 MB

Cite this

@conference{31a10b82f4284b9fb3c462e36662b0ff,

title = "Graph-Informed Neural Networks",

abstract = "Graph-Informed Neural Networks (GINNs) present a strategy for incorporating domain knowledge into scientific machine learning for complex physical systems. The construction utilizes probabilistic graphical models (PGMs) to incorporate expert knowledge, available data, constraints, etc. with physics-based models such as systems of ordinary and partial differential equations (ODEs and PDEs). Computationally intensive nodes in this hybrid model are replaced by the hidden nodes of a neural network (i.e., learned features). Once trained, the resulting GINN surrogate can cheaply generate physically-relevant predictions at scale thereby enabling robust sensitivity analysis and uncertainty quantification (UQ). As proof of concept, we build a GINN for a multiscale model of electrical double-layer capacitor dynamics embedded into a Bayesian network (BN) PDE hybrid model.",

author = "S{\o}ren Taverniers and Hall, \{Eric Joseph\} and Katsoulakis, \{Markos A.\} and Tartakovsky, \{Daniel M.\}",

note = "Funding Information: This work was performed while S. T. was employed by Stanford University. Publisher Copyright: Copyright {\textcopyright} 2021for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0); AAAI 2021 Spring Symposium Series : Combining Artificial Intelligence and Machine Learning with Physics Sciences, AAAI-MLPS 2021 ; Conference date: 22-03-2021 Through 24-03-2021",

year = "2021",

month = mar,

language = "English",

url = "https://sites.google.com/view/aaai-mlps",

}

TY - CONF

T1 - Graph-Informed Neural Networks

AU - Taverniers, Søren

AU - Hall, Eric Joseph

AU - Katsoulakis, Markos A.

AU - Tartakovsky, Daniel M.

N1 - Funding Information: This work was performed while S. T. was employed by Stanford University. Publisher Copyright: Copyright © 2021for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0)

PY - 2021/3

Y1 - 2021/3

N2 - Graph-Informed Neural Networks (GINNs) present a strategy for incorporating domain knowledge into scientific machine learning for complex physical systems. The construction utilizes probabilistic graphical models (PGMs) to incorporate expert knowledge, available data, constraints, etc. with physics-based models such as systems of ordinary and partial differential equations (ODEs and PDEs). Computationally intensive nodes in this hybrid model are replaced by the hidden nodes of a neural network (i.e., learned features). Once trained, the resulting GINN surrogate can cheaply generate physically-relevant predictions at scale thereby enabling robust sensitivity analysis and uncertainty quantification (UQ). As proof of concept, we build a GINN for a multiscale model of electrical double-layer capacitor dynamics embedded into a Bayesian network (BN) PDE hybrid model.

AB - Graph-Informed Neural Networks (GINNs) present a strategy for incorporating domain knowledge into scientific machine learning for complex physical systems. The construction utilizes probabilistic graphical models (PGMs) to incorporate expert knowledge, available data, constraints, etc. with physics-based models such as systems of ordinary and partial differential equations (ODEs and PDEs). Computationally intensive nodes in this hybrid model are replaced by the hidden nodes of a neural network (i.e., learned features). Once trained, the resulting GINN surrogate can cheaply generate physically-relevant predictions at scale thereby enabling robust sensitivity analysis and uncertainty quantification (UQ). As proof of concept, we build a GINN for a multiscale model of electrical double-layer capacitor dynamics embedded into a Bayesian network (BN) PDE hybrid model.

UR - https://www.scopus.com/pages/publications/85116636275

UR - http://CEUR-WS.org

M3 - Paper

T2 - AAAI 2021 Spring Symposium Series

Y2 - 22 March 2021 through 24 March 2021

ER -

Graph-Informed Neural Networks

Abstract

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this