[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127377-en":3,"doc-seo-127377-105":30,"detail-sidebar-cat-0-en-105":92},{"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":20,"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},127377,962085564549,"Genevieve","https://ap-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",8,"Research & Report","Multiple testing for signal-agnostic searches of new physics with machine learning","This work enhances signal-agnostic searches by introducing multiple testing strategies built on machine-learning hypothesis tests. Model selection from learning-based procedures can bias sensitivity toward particular new-physics signal families, motivating controlled combinations of tests with distinct hyperparameter choices. Results show performance comparable to the best single test while producing a more uniform response across different anomaly types. The study focuses on the New Physics Learning Machine likelihood-ratio test and evaluates methods such as p-value combination and test-statistic aggregation.","arXiv :2408 . 12296v1 [hep-ph] 22 Aug 2024  \nMultiple testing for signal-agnostic searches of new physics  \nwith machine learning  \nGaia Grosso 1,2,3 * and Marco Letizia 4,5†  \n1NSF AI Institute for Artificial Intelligence and Fundamental Interactions  \n2MIT Laboratory for Nuclear Science, Cambridge, MA  \n3School of Engineering and Applied Sciences, Harvard University, Cambridge, MA  \n4MaLGa–DIBRIS, University of Genoa, Genoa, Italy  \n5INFN, Sezione di Genova, Genoa, Italy  \nAbstract  \nIn this work, we address the question of how to enhance signal-agnostic searches by leveraging multiple testing strategies. Specifically, we consider hypothesis tests relying on machine learning, where model selection can introduce a bias towards specific families of new physics signals. We show that it is beneficial to combine different tests, characterised by distinct choices of hyperparameters, and that performances comparable to the best available test are generally achieved while providing a more uniform response to various types of anomalies. Focusing on the New Physics Learning Machine, a methodology to perform a signal-agnostic likelihood-ratio test, we explore a number of approaches to multiple testing, such as combining p-values and aggregating test statistics.  \n1. Introduction  \nAfter decades of experimental results that contributed to the development and confirmation of the Standard Model of particle physics (SM), we are in a phase in which no compelling theoretical prediction is guiding experimental searches. It is then important to develop model-independent analyses that are potentially sensitive to new physics effects not necessarily predicted by any specific Beyond the Standard Model (BSM) scenario. This is an extraordinary difficult task given the complexity of collider data and the fact that new physics can manifest itself as a deviation from the SM predictions in infinitely many ways. Moreover, these effects are expected to be extremely rare (poor signal-tobackground ratio) and/or hidden (characterising uncommon observables) .  \nSeveral proposals for partial model-independent analyses have been applied to experimental data. Early instances, such as those in [1, 2, 3, 4, 5], were based on simplifying assumptions about the way new physics effects could appear in the measurements and, as a consequence, they were limited to a selection of interesting final states. Crucially, these methodologies focused on theoretically motivated high-level features to reduce dimensionality and resorted to traditional statistical techniques.  \nRecently, machine learning has been leveraged to design flexible and multivariate data-driven tests, further enhancing signal-agnostic strategies. Various approaches have been proposed over  \n* [gaiag795@mit.edu](gaiag795@mit.edu)  \n†[marco.letizia@edu.unige.it](marco.letizia@edu.unige.it)  \nthe past few years (the reader can find an exhaustive review in [6]), some of which have already been applied to experimental data (see [7] and [8]) . Despite their potential, the adoption of these techniques introduces new challenges, particularly in understanding how model selection can impact sensitivity and bias the analysis towards specific signal hypotheses.  \nHere, we address this topic considering as a case study the model introduced in [9], where classifiers based on efficient kernel methods [10] are used to design a multivariate and unbinned likelihood-ratio test in which the alternative hypothesis is derived from data. This idea (dubbed The New Physics Learning Machine, NPLM for brevity) was initially presented in [11] using neural networks. The approach to hyperparameters selection proposed in [9] is based on a mix of statistical and heuristic criteria that was shown to work well empirically on a number of benchmarks (see also [12, 13]) . However, it is not guaranteed that the resulting model has optimal performance or a uniform response over a wide range of possible deviations from the reference expectation.","cbCaitcfjC7T3n6U","https://ap.wps.com/l/cbCaitcfjC7T3n6U","pdf",501854,1,17,"English","en",105,"# Introduction\n## Signal-agnostic analyses and machine learning\n## Motivation for multiple testing\n# The search for new physics as a signal-agnostic hypothesis test\n## Statistical framework and reference model","[{\"question\":\"Why can machine-learning model selection bias signal-agnostic searches?\",\"answer\":\"Learning-based hypothesis tests require selecting hyperparameters, and this selection can favor specific families of new-physics signals, creating sensitivity bias toward particular anomaly types.\"},{\"question\":\"How does the proposed approach improve robustness in signal-agnostic searches?\",\"answer\":\"By defining multiple tests with different hyperparameter choices (implicitly defining different alternative hypotheses) and combining their outputs into a single meta-analysis while accounting for the look-elsewhere effect.\"},{\"question\":\"Which multiple-testing integration methods are studied for the New Physics Learning Machine?\",\"answer\":\"The paper explores approaches such as combining p-values and aggregating test statistics within the signal-agnostic likelihood-ratio testing framework.\"}]","Multiple testing for signal-agnostic searches of new physics with machine learning | PDF",1785938579,43,{"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":87,"head_meta":89,"extra_data":91,"updated_unix":28},"multiple-testing-for-signal-agnostic-searches-of-new-physics-with-machine-learning","",{"@graph":36,"@context":86},[37,54,69],{"@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/multiple-testing-for-signal-agnostic-searches-of-new-physics-with-machine-learning/127377/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-23","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Why can machine-learning model selection bias signal-agnostic searches?","Question",{"text":76,"@type":77},"Learning-based hypothesis tests require selecting hyperparameters, and this selection can favor specific families of new-physics signals, creating sensitivity bias toward particular anomaly types.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How does the proposed approach improve robustness in signal-agnostic searches?",{"text":81,"@type":77},"By defining multiple tests with different hyperparameter choices (implicitly defining different alternative hypotheses) and combining their outputs into a single meta-analysis while accounting for the look-elsewhere effect.",{"name":83,"@type":74,"acceptedAnswer":84},"Which multiple-testing integration methods are studied for the New Physics Learning Machine?",{"text":85,"@type":77},"The paper explores approaches such as combining p-values and aggregating test statistics within the signal-agnostic likelihood-ratio testing framework.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]