journal article·2025-01-09Open access

Cost-Effective Strategy of Enhancing Machine Learning Potentials by Transfer Learning from a Multicomponent Data Set on ænet-PyTorch

An Niza El Aisnada, Kajjana Boonpalit, Robin van der Kruit, Koen M. Draijer, Jon López-Zorrilla, Masahiro Miyauchi, Akira Yamaguchi, Nongnuch Artrith

Abstract

High Resolution Image Download MS PowerPoint Slide Machine learning potentials (MLPs) offer efficient and accurate material simulations, but constructing the reference ab initio database remains a significant challenge, particularly for catalyst-adsorbate systems. Training an MLP with a small data set can lead to overfitting, thus limiting its practical applications. This study explores the feasibility of developing computationally cost-effective and accurate MLPs for catalyst-adsorbate systems with a limited number of ab initio references by leveraging a transfer learning strategy from subsets of a comprehensive public database. Using the Open Catalyst Project 2020 (OC20)─a data set closely related to our system of interest─we pretrained MLP models on OC20 subsets using the ænet-PyTorch framework. We compared several strategies for database subset selection. Our findings indicate that MLPs constructed via transfer learning exhibit better generalizability than those constructed from scratch, as demonstrated by the consistency in the dynamics simulations. Remarkably, transfer learning enhances the stability and accuracy of MLPs for the CuAu/H 2 O system with approximately 600 reference data points. This approach achieved excellent extrapolation performance in molecular dynamics simulations for the larger CuAu/6H 2 O system, maintaining stable and accurate predictions for up to 250 ps, whereas MLPs without transfer learning become unstable before reaching 50 ps. We also examine the potential limitations of this strategy. This work proposes an alternative, cost-effective approach for constructing MLPs for the challenging simulation of catalytic systems. Finally, we anticipate that this methodology will pave the way for broader applications in materials science and catalysis research, facilitating more efficient and accurate simulations across various systems.
Source: openalex_inverted_index · Quality: baseline

Research context

OpenAlex topics: Machine Learning in Materials Science · Computational Drug Discovery Methods · Fault Detection and Control Systems

SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.

Affiliations

Department of Materials Science and Engineering, School of Materials and Chemical Technology, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8552, Japan

Materials Chemistry and Catalysis, Debye Institute for Nanomaterials Science

Utrecht University

School of Information Science and Technology

Vidyasirimedhi Institute of Science and Technology

Physics Department

University of the Basque Country (UPV/EHU)

Biofunctional Catalyst Research Team, RIKEN Center for Sustainable Resource Science, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan

Funding

Ministry of Education, Culture, Sports, Science and Technology · Universiteit Utrecht · Nederlandse Organisatie voor Wetenschappelijk Onderzoek

References (29)

Show source-provided references
1. 10.1021/acs.jpca.2c0677810.1021/acs.jpca.2c06778 ↗
2. 10.1063/1.512633610.1063/1.5126336 ↗
3. 10.1103/PhysRevB.73.11543110.1103/PhysRevB.73.115431 ↗
4. 10.1103/PhysRevLett.98.14640110.1103/PhysRevLett.98.146401 ↗
5. 10.1038/s43588-022-00349-310.1038/s43588-022-00349-3 ↗
6. 10.1016/j.matt.2020.09.01210.1016/j.matt.2020.09.012 ↗
7. 10.1016/j.cpc.2019.04.01410.1016/j.cpc.2019.04.014 ↗
8. 10.1038/s41524-021-00559-910.1038/s41524-021-00559-9 ↗
9. 10.1016/j.cpc.2018.03.01610.1016/j.cpc.2018.03.016 ↗
10. 10.1038/s41524-020-0323-810.1038/s41524-020-0323-8 ↗
11. 10.1103/PhysRevB.96.01411210.1103/PhysRevB.96.014112 ↗
12. 10.1016/j.commatsci.2015.11.04710.1016/j.commatsci.2015.11.047 ↗
13. 10.1021/acsami.3c1539910.1021/acsami.3c15399 ↗
14. 10.1021/acs.jctc.4c0047910.1021/acs.jctc.4c00479 ↗
15. 10.1063/5.004668910.1063/5.0046689 ↗
16. 10.1021/acscatal.0c0452510.1021/acscatal.0c04525 ↗
17. 10.1063/5.014680310.1063/5.0146803 ↗
18. 10.1063/5.006388010.1063/5.0063880 ↗
19. 10.1002/ange.20170311410.1002/ange.201703114 ↗
20. 10.1103/PhysRevB.83.15310110.1103/PhysRevB.83.153101 ↗
21. 10.1063/1.505577210.1063/1.5055772 ↗
22. 10.1088/2632-2153/abdaf710.1088/2632-2153/abdaf7 ↗
23. Cheng H. · 2015 · 15510.1007/978-1-4471-6714-3_6 ↗
24. Frederic R. · 2024 · 131
25. 10.1063/5.008076610.1063/5.0080766 ↗
26. 10.1039/C6SC05720A10.1039/C6SC05720A ↗
27. 10.48550/arXiv.1711.0510110.48550/arXiv.1711.05101 ↗
28. 10.1039/D4TA03030C10.1039/D4TA03030C ↗
29. Cohen J. · 1998