{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,23]],"date-time":"2026-04-23T22:49:31Z","timestamp":1776984571635,"version":"3.51.4"},"reference-count":47,"publisher":"IOP Publishing","issue":"2","license":[{"start":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T00:00:00Z","timestamp":1770940800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T00:00:00Z","timestamp":1770940800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"funder":[{"name":"Laboratory Directed Research and Development (LDRD) Program at LLNL","award":["23-SI-006"],"award-info":[{"award-number":["23-SI-006"]}]},{"DOI":"10.13039\/100006227","name":"Lawrence Livermore National Laboratory","doi-asserted-by":"crossref","award":["DE-AC52-07NA27344"],"award-info":[{"award-number":["DE-AC52-07NA27344"]}],"id":[{"id":"10.13039\/100006227","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2026,4,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Constructing a chemically diverse dataset while avoiding sampling bias is critical to training efficient and generalizable force fields. However, in computational chemistry and materials science, many common dataset generation techniques are prone to oversampling regions of the potential energy surface. Furthermore, these regions can be difficult to identify and isolate from each other or may not align well with human intuition, making it challenging to systematically remove bias in the dataset. While traditional clustering and pruning (down-sampling) approaches can be useful for this, they can often lead to information loss or a failure to properly identify distinct regions of the potential energy surface due to difficulties associated with the high dimensionality of atomic descriptors. In this work, we introduce the Multi-kernel Edge Attention-based Graph Autoencoder (MEAGraph) model, an unsupervised approach for analyzing atomic datasets. MEAGraph combines multiple linear kernel transformations with attention-based message passing to capture geometric sensitivity and enable effective dataset pruning without relying on labels or extensive training. Demonstrated applications on niobium, tantalum, and iron datasets show that MEAGraph efficiently groups similar atomic environments, allowing for the use of basic pruning techniques for removing sampling bias. This approach provides an effective method for representation learning and clustering that can be used for data analysis, outlier detection, and dataset optimization.<\/jats:p>","DOI":"10.1088\/2632-2153\/ae39a1","type":"journal-article","created":{"date-parts":[[2026,1,16]],"date-time":"2026-01-16T22:53:37Z","timestamp":1768604017000},"page":"025003","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Unsupervised atomic data mining via multi-kernel graph autoencoders for machine learning force fields"],"prefix":"10.1088","volume":"7","author":[{"given":"Hong","family":"Sun","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9191-055X","authenticated-orcid":true,"given":"Joshua A","family":"Vita","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Amit","family":"Samanta","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2415-4656","authenticated-orcid":false,"given":"Vincenzo","family":"Lordi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"266","published-online":{"date-parts":[[2026,2,13]]},"reference":[{"key":"mlstae39a1bib1","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevLett.98.146401","type":"journal-article","article-title":"Generalized neural-network representation of high-dimensional potential-energy surfaces","volume":"98","author":"Behler","year":"2007","journal-title":"Phys. Rev. Lett."},{"key":"mlstae39a1bib2","doi-asserted-by":"publisher","DOI":"10.1038\/ncomms13890","type":"journal-article","article-title":"Quantum-chemical insights from deep tensor neural networks","volume":"8","author":"Sch\u00fctt","year":"2017","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib3","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevLett.104.136403","type":"journal-article","article-title":"Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons","volume":"104","author":"Bart\u00f3k","year":"2010","journal-title":"Phys. Rev. Lett."},{"key":"mlstae39a1bib4","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.1701816","type":"journal-article","article-title":"Machine learning unifies the modeling of materials and molecules","volume":"3","author":"Bart\u00f3k","year":"2017","journal-title":"Sci. Adv."},{"key":"mlstae39a1bib5","doi-asserted-by":"publisher","first-page":"677","DOI":"10.1021\/acs.chemrev.1c00022","type":"journal-article","article-title":"Gaussian process regression for materials and molecules","volume":"6","author":"Deringer","year":"2021","journal-title":"Nat. Rev. Mater."},{"key":"mlstae39a1bib6","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevLett.108.058301","type":"journal-article","article-title":"Fast and accurate modeling of molecular atomization energies with machine learning","volume":"108","author":"Rupp","year":"2012","journal-title":"Phys. Rev. Lett."},{"key":"mlstae39a1bib7","doi-asserted-by":"publisher","DOI":"10.1088\/0953-8984\/26\/18\/183001","type":"journal-article","article-title":"Representing potential energy surfaces by high-dimensional neural network potentials","volume":"26","author":"Behler","year":"2014","journal-title":"J. Phys.: Condens. Matter"},{"key":"mlstae39a1bib8","doi-asserted-by":"publisher","first-page":"22790","DOI":"10.1021\/acs.jpcc.8b08063","type":"journal-article","article-title":"Toward reliable and transferable machine learning potentials: uniform training by overcoming sampling bias","volume":"122","author":"Jeong","year":"2018","journal-title":"J. Phys. Chem. C"},{"key":"mlstae39a1bib9","doi-asserted-by":"publisher","first-page":"13681","DOI":"10.1021\/acs.chemrev.4c00572","type":"journal-article","article-title":"Data generation for machine learning interatomic potentials and beyond","volume":"124","author":"Kulichenko","year":"2024","journal-title":"Chem. Rev."},{"key":"mlstae39a1bib10","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1038\/s41524-024-01500-6","type":"journal-article","article-title":"Systematic softening in universal machine learning interatomic potentials","volume":"11","author":"Deng","year":"2025","journal-title":"npj Comput. Mater."},{"key":"mlstae39a1bib11","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2511.10561","type":"other","article-title":"Maximizing efficiency of dataset compression for machine learning potentials with information theory","author":"Yu","year":"2025"},{"key":"mlstae39a1bib12","doi-asserted-by":"publisher","first-page":"19523","DOI":"10.48550\/arXiv.2206.14486","type":"conference-proceedings","article-title":"Beyond neural scaling laws: beating power law scaling via data pruning","volume":"vol 35","author":"Sorscher","year":"2022"},{"key":"mlstae39a1bib13","doi-asserted-by":"publisher","first-page":"21455","DOI":"10.48550\/arXiv.2208.05516","type":"conference-proceedings","article-title":"Quality not quantity: on the interaction between dataset design and robustness of clip","volume":"vol 35","author":"Nguyen","year":"2022"},{"key":"mlstae39a1bib14","doi-asserted-by":"publisher","first-page":"9075","DOI":"10.48550\/arXiv.2006.10621","type":"conference-proceedings","article-title":"On the predictability of pruning across scales","author":"Rosenfeld","year":"2021"},{"key":"mlstae39a1bib15","doi-asserted-by":"publisher","first-page":"2818","DOI":"10.1109\/CVPR52729.2023.00276","type":"conference-proceedings","article-title":"Reproducible scaling laws for contrastive language-image learning","author":"Cherti","year":"2023"},{"key":"mlstae39a1bib16","doi-asserted-by":"publisher","first-page":"53983","DOI":"10.48550\/arXiv.2308.12284","type":"conference-proceedings","article-title":"D4: improving llm pretraining via document de-duplication and diversification","volume":"vol 36","author":"Tirumala","year":"2023"},{"key":"mlstae39a1bib17","doi-asserted-by":"publisher","first-page":"236","DOI":"10.18653\/v1\/2023.findings-emnlp.18","type":"book","article-title":"Data pruning for efficient model pruning in neural machine translation","author":"Azeemi","year":"2023"},{"key":"mlstae39a1bib18","doi-asserted-by":"publisher","first-page":"131","DOI":"10.1038\/s41524-021-00595-5","type":"journal-article","article-title":"Metadynamics sampling in atomic environment space for collecting training data for machine learning potentials","volume":"7","author":"Yoo","year":"2021","journal-title":"npj Comput. Mater."},{"key":"mlstae39a1bib19","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1812.05159","type":"conference-proceedings","article-title":"An empirical study of example forgetting during deep neural network learning","author":"Toneva","year":"2019"},{"key":"mlstae39a1bib20","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2107.07075","type":"conference-proceedings","article-title":"Deep learning on a data diet: finding important examples early in training","author":"Paul","year":"2022"},{"key":"mlstae39a1bib21","doi-asserted-by":"publisher","first-page":"189","DOI":"10.1038\/s41524-022-00872-x","type":"journal-article","article-title":"Training data selection for accuracy and transferability of interatomic potentials","volume":"8","author":"de Oca Zapiain","year":"2022","journal-title":"npj Comput. Mater."},{"key":"mlstae39a1bib22","doi-asserted-by":"publisher","first-page":"4014","DOI":"10.1038\/s41467-025-59232-0","type":"journal-article","article-title":"Model-free estimation of completeness, uncertainties and outliers in atomistic machine learning using information theory","volume":"16","author":"Schwalbe-Koda","year":"2025","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib23","doi-asserted-by":"publisher","first-page":"230","DOI":"10.1038\/s43588-023-00406-5","type":"journal-article","article-title":"Uncertainty-driven dynamics for active learning of interatomic potentials","volume":"3","author":"Kulichenko","year":"2023","journal-title":"Nat. Comput. Sci."},{"key":"mlstae39a1bib24","doi-asserted-by":"publisher","first-page":"5104","DOI":"10.1038\/s41467-021-25342-8","type":"journal-article","article-title":"Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks","volume":"12","author":"Schwalbe-Koda","year":"2021","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib25","doi-asserted-by":"publisher","first-page":"605","DOI":"10.1039\/D2DD00034B","type":"journal-article","article-title":"Exploring chemical and conformational spaces by batch mode deep active learning","volume":"1","author":"Zaverkin","year":"2022","journal-title":"Digit. Discovery"},{"key":"mlstae39a1bib26","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevB.87.184115","type":"journal-article","article-title":"On representing chemical environments","volume":"87","author":"Bart\u00f3k","year":"2013","journal-title":"Phys. Rev. B"},{"key":"mlstae39a1bib27","doi-asserted-by":"publisher","DOI":"10.1063\/1.5019667","type":"journal-article","article-title":"wacsf-weighted atom-centered symmetry functions as descriptors in machine learning potentials","volume":"148","author":"Gastegger","year":"2018","journal-title":"J. Chem. Phys."},{"key":"mlstae39a1bib28","doi-asserted-by":"publisher","first-page":"5118","DOI":"10.21105\/joss.05118","type":"journal-article","article-title":"Fitsnap: atomistic machine learning with lammps","volume":"8","author":"Rohskopf","year":"2023","journal-title":"J. Open Source Softw."},{"key":"mlstae39a1bib29","doi-asserted-by":"publisher","first-page":"2453","DOI":"10.1038\/s41467-022-29939-5","type":"journal-article","article-title":"E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials","volume":"13","author":"Batzner","year":"2022","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib30","doi-asserted-by":"publisher","first-page":"11423","DOI":"10.48550\/arXiv.2206.07697","type":"conference-proceedings","article-title":"Mace: Higher order equivariant message passing neural networks for fast and accurate force fields","volume":"vol 35","author":"Batatia","year":"2022"},{"key":"mlstae39a1bib31","doi-asserted-by":"publisher","first-page":"53","DOI":"10.1016\/j.patrec.2015.04.009","type":"journal-article","article-title":"What are the true clusters?","volume":"64","author":"Hennig","year":"2015","journal-title":"Pattern Recognit. Lett."},{"key":"mlstae39a1bib32","doi-asserted-by":"publisher","first-page":"426","DOI":"10.1590\/s1982-21702018000400027","type":"journal-article","article-title":"Evaluation of the clustering performance of affinity propagation algorithm considering the influence of preference parameter and damping factor","volume":"24","author":"Fenias Moiane","year":"2018","journal-title":"Boletim de Ci\u00eancias Geod\u00e9sicas"},{"key":"mlstae39a1bib33","doi-asserted-by":"publisher","DOI":"10.1109\/MFI49285.2020.9235263","type":"other","article-title":"A hybrid approach to hierarchical density-based cluster selection","author":"Malzer","year":"2019"},{"key":"mlstae39a1bib34","article-title":"fitsnap Ta","author":"FitSNAP Development Team","year":"2014","type":"other"},{"key":"mlstae39a1bib35","article-title":"fitsnap Fe","author":"FitSNAP Development Team","year":"2014","type":"other"},{"key":"mlstae39a1bib36","doi-asserted-by":"publisher","DOI":"10.1016\/j.commatsci.2023.112497","type":"journal-article","article-title":"Exploring structural transitions at grain boundaries in Nb using a generalized embedded atom interatomic potential","volume":"230","author":"Sun","year":"2023","journal-title":"Comput. Mater. Sci."},{"key":"mlstae39a1bib37","doi-asserted-by":"publisher","first-page":"784","DOI":"10.1103\/PhysRevB.28.784","type":"journal-article","article-title":"Bond-orientational order in liquids and glasses","volume":"28","author":"Steinhardt","year":"1983","journal-title":"Phys. Rev. B"},{"key":"mlstae39a1bib38","doi-asserted-by":"publisher","first-page":"6056","DOI":"10.1038\/s41467-021-26199-7","type":"journal-article","article-title":"Data-driven simulation and characterisation of gold nanoparticle melting","volume":"12","author":"Zeni","year":"2021","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib39","doi-asserted-by":"publisher","first-page":"5461","DOI":"10.1038\/s41467-020-19168-z","type":"journal-article","article-title":"A general-purpose machine-learning force field for bulk and nanostructured phosphorus","volume":"11","author":"Deringer","year":"2020","journal-title":"Nat. Commun."},{"key":"mlstae39a1bib40","doi-asserted-by":"publisher","DOI":"10.1111\/jace.20128","type":"journal-article","article-title":"Multikernel similarity-based clustering of amorphous systems and machine-learned interatomic potentials by active learning","volume":"108","author":"Shuaib","year":"2024","journal-title":"J. Am. Ceram. Soc."},{"key":"mlstae39a1bib41","doi-asserted-by":"publisher","first-page":"232","DOI":"10.1109\/ICADIWT.2014.6814687","type":"conference-proceedings","article-title":"Dbscan: past, present and future","author":"Khan","year":"2014"},{"key":"mlstae39a1bib42","doi-asserted-by":"publisher","first-page":"451","DOI":"10.1016\/S0031-3203(02)00060-2","type":"journal-article","article-title":"The global k-means clustering algorithm","volume":"36","author":"Likas","year":"2003","journal-title":"Pattern Recognit."},{"key":"mlstae39a1bib43","doi-asserted-by":"publisher","first-page":"395","DOI":"10.1007\/s11222-007-9033-z","type":"journal-article","article-title":"A tutorial on spectral clustering","volume":"17","author":"Luxburg","year":"2007","journal-title":"Stat. Comput."},{"key":"mlstae39a1bib44","article-title":"Adaptive affinity propagation clustering","author":"Wang","year":"2008","type":"preprint"},{"key":"mlstae39a1bib45","doi-asserted-by":"publisher","first-page":"50","DOI":"10.1038\/s41524-025-01544-2","type":"journal-article","article-title":"Unsupervised identification of crystal defects from atomistic potential descriptors","volume":"11","author":"K\u00fdvala","year":"2025","journal-title":"npj Comput. Mater."},{"key":"mlstae39a1bib46","doi-asserted-by":"publisher","first-page":"8060","DOI":"10.1021\/acs.jctc.5c00449","type":"journal-article","article-title":"Opportunities and challenges in unsupervised learning: The case of aqueous electrolyte solutions","volume":"21","author":"Sormani","year":"2025","journal-title":"J. Chem. Theory Comput."},{"key":"mlstae39a1bib47","type":"journal-article","article-title":"MEAGraph: multi-kernel edge attention graph autoencoder","author":"Sun","year":"2024","journal-title":"Github"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T15:20:22Z","timestamp":1770996022000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/ae39a1"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,13]]},"references-count":47,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2026,2,13]]},"published-print":{"date-parts":[[2026,4,1]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/ae39a1","relation":{},"ISSN":["2632-2153"],"issn-type":[{"value":"2632-2153","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,13]]},"assertion":[{"value":"Unsupervised atomic data mining via multi-kernel graph autoencoders for machine learning force fields","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2026 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2025-09-15","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-01-16","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2026-02-13","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}