Learning atomic forces from uncertainty-calibrated adversarial attacks
Source abstract
Abstract
Abstract Adversarial approaches, which intentionally challenge machine learning models by generating difficult examples, are increasingly being adopted to improve machine learning interatomic potentials (MLIPs). While already providing great practical value, little is known about the actual prediction errors of MLIPs on adversarial structures and whether these errors can be controlled. We propose the Calibrated Adversarial Geometry Optimization (CAGO) algorithm to discover adversarial structures with user-assigned errors. Through uncertainty calibration, the estimated uncertainty of MLIPs is unified with real errors. By performing geometry optimization for calibrated uncertainty, we reach adversarial structures with the user-assigned target MLIP prediction error. Integrating with active learning pipelines, we benchmark CAGO, demonstrating stable MLIPs that systematically converge structural, dynamical, and thermodynamical properties for liquid water and water adsorption in a metal-organic framework within only hundreds of training structures, where previously many thousands were typically required.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Machine Learning in Materials Science · Nuclear Physics and Applications · Force Microscopy Techniques and Applications
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (42)
Show source-provided references
1. Behler, J. & Parrinello, M. Generalized neural-network representation of high-dimensional potential-energy surfaces. Phys. Rev. Lett. 98, 146401 (2007).10.1103/PhysRevLett.98.146401 ↗
2. Batzner, S. et al. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nat. Commun. 13, 2453 (2022).10.1038/s41467-022-29939-5 ↗
3. Musaelian, A. et al. Learning local equivariant representations for large-scale atomistic dynamics. Nat. Commun. 14, 579 (2023).10.1038/s41467-023-36329-y ↗
4. Batatia, I., Kovacs, D. P., Simm, G., Ortner, C. & Csanyi, G. Mace: Higher order equivariant message passing neural networks for fast and accurate force fields. In Koyejo, S. et al. (eds.) Advances in Neural Information Processing Systems, vol. 35, 11423–11436 (Curran Associates, Inc., 2022).
5. Zhai, Y., Caruso, A., Bore, S. L., Luo, Z. & Paesani, F. A “short blanket” dilemma for a state-of-the-art neural network potential for water: Reproducing experimental properties or the physics of the underlying many-body interactions? J. Chem. Phys. 158, 084111 (2023).10.1063/5.0142843 ↗
6. Fu, X. et al. Forces are not enough: Benchmark and critical evaluation for machine learning force fields with molecular simulations. Transactions on Machine Learning Research (2023).
7. Shimodaira, H. Improving predictive inference under covariate shift by weighting the log-likelihood function. J. Stat. Plan. Inference 90, 227–244 (2000).10.1016/S0378-3758(00)00115-4 ↗
8. Yang, Y., Zhang, S., Ranasinghe, K. D., Isayev, O. & Roitberg, A. E. Machine learning of reactive potentials. Annu. Rev. Phys. Chem. 75, 371–395 (2024).10.1146/annurev-physchem-062123-024417 ↗
9. Vandermause, J. et al. On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events. npj Comput. Mater. 6, 20 (2020).10.1038/s41524-020-0283-z ↗
10. Xie, Y. et al. Uncertainty-aware molecular dynamics from Bayesian active learning for phase transformations and thermal transport in SiC. npj Comput. Mater. 9, 36 (2023).10.1038/s41524-023-00988-8 ↗
11. Wen, M. & Tadmor, E. B. Uncertainty quantification in molecular simulations with dropout neural network potentials. npj Comput. Mater. 6, 124 (2020).10.1038/s41524-020-00390-8 ↗
12. Krogh, A. & Vedelsby, J. Neural network ensembles, cross validation, and active learning. In Tesauro, G., Touretzky, D. & Leen, T. (eds.) Advances in Neural Information Processing Systems, vol. 7 (MIT Press, 1994).
13. Smith, J. S., Nebgen, B., Lubbers, N., Isayev, O. & Roitberg, A. E. Less is more: Sampling chemical space with active learning. J. Chem. Phys. 148, 241733 (2018).10.1063/1.5023802 ↗
14. Zhang, Y. et al. DP-GEN: A concurrent learning platform for the generation of reliable deep learning based potential energy models. Computer Phys. Commun. 253, 107206 (2020).10.1016/j.cpc.2020.107206 ↗
15. Schran, C. et al. Machine learning potentials for complex aqueous systems made simple. Proc. Natl. Acad. Sci. USA 118, e2110077118 (2021).10.1073/pnas.2110077118 ↗
16. Vandenhaute, S., Cools-Ceuppens, M., DeKeyser, S., Verstraelen, T. & Van Speybroeck, V. Machine learning potentials for metal-organic frameworks using an incremental learning approach. npj Comput. Mater. 9, 19 (2023).10.1038/s41524-023-00969-x ↗
17. Yang, M., Bonati, L., Polino, D. & Parrinello, M. Using metadynamics to build neural network potentials for reactive events: the case of urea decomposition in water. Catal. Today 387, 143–149 (2022).10.1016/j.cattod.2021.03.018 ↗
18. Smith, J. S., Nebgen, B., Lubbers, N., Isayev, O. & Roitberg, A. E. Less is more: Sampling chemical space with active learning. J. Chem. Phys. 148 (2018).10.1063/1.5023802 ↗
19. Zeng, J., Cao, L., Xu, M., Zhu, T. & Zhang, J. Z. Complex reaction processes in combustion unraveled by neural network-based molecular dynamics simulation. Nat. Commun. 11, 5713 (2020).10.1038/s41467-020-19497-z ↗
20. Goodfellow, I. Deep learning, vol. 196 (MIT press, 2016).
21. Cubuk, E. D. & Schoenholz, S. S. Adversarial Forces of Physical Models. 3rd NeurIPS workshop on Machine Learning and the Physical Sciences (2020).
22. Kulichenko, M. et al. Uncertainty-driven dynamics for active learning of interatomic potentials. Nat. Comput. Sci. 3, 230–239 (2023).10.1038/s43588-023-00406-5 ↗
23. Schwalbe-Koda, D., Tan, A. R. & Gómez-Bombarelli, R. Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks. Nat. Commun. 12, 5104 (2021).10.1038/s41467-021-25342-8 ↗
24. Roy, S., Dürholt, J. P., Asche, T. S., Zipoli, F. & Gómez-Bombarelli, R. Learning a reactive potential for silica-water through uncertainty attribution. Nat. Commun. 15, 6030 (2024).10.1038/s41467-024-50407-9 ↗
25. Zaverkin, V. et al. Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials. npj Comput. Mater. 10, 83 (2024).10.1038/s41524-024-01254-1 ↗
26. Palmer, G. et al. Calibration after bootstrap for accurate uncertainty quantification in regression models. npj Comput. Mater. 8, 115 (2022).10.1038/s41524-022-00794-8 ↗
27. Musil, F., Willatt, M. J., Langovoy, M. A. & Ceriotti, M. Fast and accurate uncertainty estimation in chemical machine learning. J. Chem. Theory Comput. 15, 906–915 (2019).10.1021/acs.jctc.8b00959 ↗
28. Imbalzano, G. et al. Uncertainty estimation for molecular dynamics and sampling. J. Chem. Phys. 154, 074102 (2021).10.1063/5.0036522 ↗
29. Bore, S. L. & Paesani, F. Realistic phase diagram of water from “first principles” data-driven quantum simulations. Nat. Commun. 14, 3349 (2023).10.1038/s41467-023-38855-1 ↗
30. Dasgupta, S., Lambros, E., Perdew, J. P. & Paesani, F. Elevating density functional theory to chemical accuracy for water simulations through a density-corrected many-body formalism. Nat. Commun. 12, 6359 (2021).10.1038/s41467-021-26618-9 ↗
31. Dasgupta, S., Cassone, G. & Paesani, F. Nuclear quantum effects and the Grotthuss mechanism dictate the pH of liquid water (2024).10.26434/chemrxiv-2024-zkz7v ↗
32. Zhai, Y., Rashmi, R., Palos, E. & Paesani, F. Many-body interactions and deep neural network potentials for water. J. Chem. Phys. 160 (2024).10.1063/5.0203682 ↗
33. Maxson, T. & Szilvási, T. Transferable water potentials using equivariant neural networks. J. Phys. Chem. Lett. 15, 3740–3747 (2024).10.1021/acs.jpclett.4c00605 ↗
34. Cavka, J. H. et al. A new zirconium inorganic building brick forming metal organic frameworks with exceptional stability. J. Am. Chem. Soc. 130, 13850–13851 (2008).10.1021/ja8057953 ↗
35. Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods 17, 261–272 (2020).10.1038/s41592-019-0686-2 ↗
36. Zeng, J. et al. DeePMD-kit v2: A software package for deep potential models. J. Chem. Phys. 159, 054801 (2023).10.1063/5.0155600 ↗
37. Sciortino, F., Zhai, Y., Bore, S. & Paesani, F. Constraints on the location of the liquid–liquid critical point in water. Nature Physics 1–6 (2025).10.26434/chemrxiv-2024-dqqws ↗
38. Babin, V., Leforestier, C. & Paesani, F. Development of a “first principles” water potential with flexible monomers: Dimer potential energy surface, vrt spectrum, and second virial coefficient. J. Chem. Theory Comput. 9, 5395–5403 (2013).10.1021/ct400863t ↗
39. Babin, V., Medders, G. R. & Paesani, F. Development of a “first principles” water potential with flexible monomers. ii: Trimer potential energy surface, third virial coefficient, and small clusters. J. Chem. Theory Comput. 10, 1599–1607 (2014).10.1021/ct500079y ↗
40. Belleflamme, F. & Hutter, J. Radicals in aqueous solution: assessment of density-corrected SCAN functional. Phys. Chem. Chem. Phys. 25, 20817–20836 (2023).10.1039/D3CP02517A ↗
41. Hylleraas Software Platformhyal, https://gitlab.com/hylleraasplatform (2025).
42. Thompson, A. P. et al. LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Phys. Commun. 271, 108171 (2022).10.1016/j.cpc.2021.108171 ↗