Guardado en:
Detalles Bibliográficos
Autores principales: Fronzi, Marco, Ford, Michael J., Nayal, Kamal Singh, Isayev, Olexandr, Stampfl, Catherine
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2510.02681
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917140464926720
author Fronzi, Marco
Ford, Michael J.
Nayal, Kamal Singh
Isayev, Olexandr
Stampfl, Catherine
author_facet Fronzi, Marco
Ford, Michael J.
Nayal, Kamal Singh
Isayev, Olexandr
Stampfl, Catherine
contents The discovery of high-performance thermoelectric materials requires models that are both accurate and interpretable. Traditional machine learning approaches, while effective at property prediction, often act as black boxes and provide limited physical insight. In this work, we introduce Kolmogorov--Arnold Networks (KANs) for the prediction of thermoelectric properties, focusing on the Seebeck coefficient and band gap. Compared to multilayer perceptrons (MLPs), KANs achieve comparable predictive accuracy while offering explicit symbolic representations of structure--property relationships. This dual capability enables both reliable predictions and the extraction of physically meaningful functional forms. Benchmarking against literature models further highlights the robustness and generalisability of the approach. Our findings demonstrate that KANs provide a powerful framework for reverse engineering materials with targeted thermoelectric properties, bridging the gap between predictive performance and scientific interpretability.
format Preprint
id arxiv_https___arxiv_org_abs_2510_02681
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Kolmogorov-Arnold Networks in Thermoelectric Materials Design
Fronzi, Marco
Ford, Michael J.
Nayal, Kamal Singh
Isayev, Olexandr
Stampfl, Catherine
Materials Science
The discovery of high-performance thermoelectric materials requires models that are both accurate and interpretable. Traditional machine learning approaches, while effective at property prediction, often act as black boxes and provide limited physical insight. In this work, we introduce Kolmogorov--Arnold Networks (KANs) for the prediction of thermoelectric properties, focusing on the Seebeck coefficient and band gap. Compared to multilayer perceptrons (MLPs), KANs achieve comparable predictive accuracy while offering explicit symbolic representations of structure--property relationships. This dual capability enables both reliable predictions and the extraction of physically meaningful functional forms. Benchmarking against literature models further highlights the robustness and generalisability of the approach. Our findings demonstrate that KANs provide a powerful framework for reverse engineering materials with targeted thermoelectric properties, bridging the gap between predictive performance and scientific interpretability.
title Kolmogorov-Arnold Networks in Thermoelectric Materials Design
topic Materials Science
url https://arxiv.org/abs/2510.02681