Scalable Training of Trustworthy and Energy-Efficient Predictive Graph Foundation Models for Atomistic Materials Modeling: A Case Study with HydraGNN

We present our work on developing and training scalable, trustworthy, and energy-efficient predictive graph foundation models (GFMs) using HydraGNN, a multi-headed graph convolutional neural network architecture. HydraGNN expands the boundaries of graph neural network (GNN) computations in both trai...

Descripción completa

Detalles Bibliográficos

Autores principales:	Pasini, Massimiliano Lupo, Choi, Jong Youl, Mehta, Kshitij, Zhang, Pei, Rogers, David, Bae, Jonghyun, Ibrahim, Khaled Z., Aji, Ashwin M., Schulz, Karl W., Polo, Jorda, Balaprakash, Prasanna
Formato:	Artículo
Publicado:	2024
Materias:	Machine Learning Computational Physics 68T07, 68T09 C.2.4; I.2.11
Acceso en línea:	https://arxiv.org/abs/2406.12909 http://arxiv.org/abs/2406.12909
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!