{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/metadynamics-for-training-neural-network","title":"Metadynamics for Training Neural Network Model Chemistries: a Competitive Assessment","arxiv_id":"1712.07240","date":"2017-12-19","proceeding":null,"authors":["John E. Herr","Kun Yao","Ryker McIntyre","David Toth","John Parkhill"],"abstract":"Neural network (NN) model chemistries (MCs) promise to facilitate the\naccurate exploration of chemical space and simulation of large reactive\nsystems. One important path to improving these models is to add layers of\nphysical detail, especially long-range forces. At short range, however, these\nmodels are data driven and data limited. Little is systematically known about\nhow data should be sampled, and `test data' chosen randomly from some sampling\ntechniques can provide poor information about generality. If the sampling\nmethod is narrow `test error' can appear encouragingly tiny while the model\nfails catastrophically elsewhere. In this manuscript we competitively evaluate\ntwo common sampling methods: molecular dynamics (MD), normal-mode sampling\n(NMS) and one uncommon alternative, Metadynamics (MetaMD), for preparing\ntraining geometries. We show that MD is an inefficient sampling method in the\nsense that additional samples do not improve generality. We also show MetaMD is\neasily implemented in any NNMC software package with cost that scales linearly\nwith the number of atoms in a sample molecule. MetaMD is a black-box way to\nensure samples always reach out to new regions of chemical space, while\nremaining relevant to chemistry near $k_bT$. It is one cheap tool to address\nthe issue of generalization.","url_abs":"http://arxiv.org/abs/1712.07240v1","url_pdf":"http://arxiv.org/pdf/1712.07240v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"metadynamics-for-training-neural-network","repo_url":"https://github.com/jparkhill/TensorMol","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}