[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119707-en":3,"doc-seo-119707-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119707,1099513958762,"Logic","https://ap-avatar.wpscdn.com/avatar/1000023916a998db790?x-image-process=image/resize,m_fixed,w_180,h_180&k=1784791008015729253",8,"Research & Report","Efficient Generation of Stable Linear Machine-Learning Force Fields with Uncertainty-Aware Active Learning","Machine-learning force fields can represent molecular and materials potential energy surfaces using limited ab initio training data, but scaling these approaches requires accurate models trained with as few calculations as possible. This work investigates an uncertainty-aware active-learning strategy that combines linear regression with uncertainty prediction for unsampled molecular configurations. Applied to spectral neighbor analysis potentials, the method generates stable force fields for room-temperature molecular dynamics using only tens of ab initio force simulations, achieving chemical accuracy near its target. It also avoids conformational pre-sampling, minimizing user intervention.","arXiv :2303 . 16538v1 [physics .comp-ph] 29 Mar 2023  \nE􀀎cient Generation of Stable Linear Machine-Learning Force Fields with Uncertainty-Aware Active Learning  \nValerio Briganti and Alessandro Lunghi 􀀃  \nSchool of Physics, AMBER and CRANN Institute, Trinity College, Dublin 2, Ireland  \nMachine-learning force 􀀌elds enable an accurate and universal description of the potential energy surface of molecules and materials on the basis of a training set of ab initio data. However, large-scale applications of these methods rest on the possibility to train accurate machine learning models with a small number of ab initio data. In this respect, active-learning strategies, where the training set is self-generated by the model itself, combined with linear machine-learning models are particularly promising. In this work, we explore an active-learning strategy based on linear regression and able to predict the model's uncertainty on predictions for molecular con􀀌gurations not sampled by the training set, thus providing a straightforward recipe for the extension of the latter. We apply this strategy to the spectral neighbor analysis potential and show that only tens of ab initio simulations of atomic forces are required to generate stable force 􀀌elds for room-temperature molecular dynamics at or close to chemical accuracy. Moreover, the method does not necessitate any conformational pre-sampling, thus requiring minimal user intervention and parametrization.  \nINTRODUCTION  \nMachine learning (ML) models for the generation of force 􀀌elds (FFs) are becoming a prominent aid for researchers in di􀀋erent 􀀌elds, including drug discovery[1], prediction of metastable structures[2], heterogenous catalysis [3], and more[4{7] . In all these 􀀌elds, ML permits to speed up calculations or to manage larger datasets, largely overcoming the problem of the computational costs inherent to electronic structure simulations. In recent years, many ML models for the generation of FFs have been presented, e.g. sGDML[8], BP-NNP[9{11], GPR based models[12{14], PhysNet[15], SchNet[16], FCHL19 descriptors combined with di􀀋erent regressors[17], moment tensor potentials[18], message passage neural networks[19{21], and many more. All these methods have been shown to be able to reproduce the potential energy surface (PES) of complex chemical systems with chemical or near-to-chemical accuracy. However, such incredible results often come with the burden of requiring a lot of electronic structure simulations to generate the necessary training data to reach high accuracy, often in the range of 103 􀀀 106 calculations[22{24] . Such a scenario poses serious challenges to the widespread use of MLFFs.  \nDecreasing the size of the training set is a non-trivial challenge that depends on many di􀀋erent factors. Among the most crucial ones there is the complexity of the ML architecture used to map the PES and the approach used to select a training set. Although simple ML models, such as linear ones, achieve less accuracy than complex ones, they often perform better for small training sets in virtue of being less prone to over-􀀌tting  \n􀀃 [lunghia@tcd.ie](lunghia@tcd.ie)  \nissues. In this work we will focus on this class of MLFFsand investigate the possibility to further optimize their generation in terms of accuracy and training set size.  \nA conventional way to learn the PES of a compound is to 􀀌rst perform ab initio molecular dynamics to sample a relevant number of con􀀌gurations and their energy/forces[9, 22] . This approach can potentially achieve a good performance on both training and test sets, as the most statically relevant structures are automatically included. However, such approach does not guarantee that redundancies are not also included, potentially leading to large computational overheads. Moreover, the accurate representation of a molecular PES also requires the sampling of statistically-rare conformations, which by de􀀌nition are not captured by small-size molecular dynamic","cbCaie4japvhooHN","https://ap.wps.com/l/cbCaie4japvhooHN","pdf",1403436,1,12,"English","en",105,"# Introduction\n## Machine-learning force fields and training-data cost\n## Reducing training set size with simpler ML models\n## Conventional ab initio sampling vs stability challenges\n## Active learning strategies for optimal training sets\n## Query by committee approach\n## Bayesian uncertainty and Gaussian Process regression","[{\"question\":\"What problem does the paper address in machine-learning force fields?\",\"answer\":\"Large-scale use requires accurate force-field models trained with very small ab initio datasets, while avoiding instability when rare configurations appear.\"},{\"question\":\"How does the proposed active-learning strategy work?\",\"answer\":\"It uses a linear regression model that can predict uncertainty for molecular configurations not included in the current training set, guiding which new configurations to add.\"},{\"question\":\"What performance is demonstrated and how many simulations are needed?\",\"answer\":\"For spectral neighbor analysis potentials, only tens of ab initio force simulations are required to generate stable force fields for room-temperature molecular dynamics at or close to chemical accuracy.\"}]","Efficient Generation of Stable Linear Machine-Learning Force Fields with Uncertainty-Aware Active Learning | PDF",1785725884,30,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"efficient-generation-of-stable-linear-machine-learning-force-fields-with-uncertainty-aware-active-learning","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/efficient-generation-of-stable-linear-machine-learning-force-fields-with-uncertainty-aware-active-learning/119707/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in machine-learning force fields?","Question",{"text":75,"@type":76},"Large-scale use requires accurate force-field models trained with very small ab initio datasets, while avoiding instability when rare configurations appear.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed active-learning strategy work?",{"text":80,"@type":76},"It uses a linear regression model that can predict uncertainty for molecular configurations not included in the current training set, guiding which new configurations to add.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance is demonstrated and how many simulations are needed?",{"text":84,"@type":76},"For spectral neighbor analysis potentials, only tens of ab initio force simulations are required to generate stable force fields for room-temperature molecular dynamics at or close to chemical accuracy.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]