[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-118161-en":3,"doc-seo-118161-105":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},118161,8796095462418,"Noah","https://ap-avatar.wpscdn.com/avatar/80000253c1241d02b47?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778826106357471780",8,"Research & Report","Ultrafast jet classification at the HL-LHC","Three machine learning models are developed for jet origin classification, with a focus on deployment on field-programmable gate arrays used in the HL-LHC Level-1 trigger. The work characterizes how inference latency and resource usage change with input size and algorithm choice, while ensuring compatibility with the expected detector data type and high-luminosity operating conditions at CERN. Using quantization-aware training and FPGA-oriented efficient synthesis, the study shows that complex architectures such as Deep Sets and Interaction Networks can reach O(100) ns inference with relatively low computational cost.","arXiv :2402 .0 1876v2 [hep-ex] 4 Jul 2024  \nFERMILAB-PUB-24-0030-CMS-CSAID-PPD  \nUltrafast jet classification at the HL-LHC  \nPatrick Odagiu 1 , Zhiqiang Que2 , Javier Duarte3 , Johannes Haller4 , Gregor Kasieczka4 , Artur Lobanov4 , Vladimir Loncar5 , 12 Wayne Luk2 , Jennifer Ngadiuba6 , Maurizio Pierini7 , Philipp Rincke8 , 10 , Arpita Seksaria 10 , Sioni Summers7 , Andre Sznajder 11 , Alexander Tapper2 , Thea K. ˚Arrestad 1  \n1 ETH Z¨urich, Z¨urich, Switzerland,  \n2 Imperial College London, London, UK,  \n3 University of California San Diego, La Jolla, CA, USA,  \n4 Universit¨at Hamburg, Hamburg, Germany,  \n5 Massachusetts Institute of Technology, Cambridge, MA, USA  \n6 Fermi National Accelerator Laboratory, Batavia, IL, USA,  \n7 European Organization for Nuclear Research (CERN), Geneva, Switzerland,  \n8 Universit¨at G¨ottingen, G¨ottingen, Germany,  \n9 University of Southern California,Los Angeles, CA, USA,  \n10 Uppsala Universitet, Uppsala, Sweden,  \n11 Universidade do Estado do Rio de Janeiro (UERJ), Rio de Janeiro, Brazil,  \n12 Institute of Physics Belgrade, Serbia E-mail: [podagiu@ethz.ch](podagiu@ethz.ch)  \n8 July 2024  \nAbstract. Three machine learning models are used to perform jet origin classification. These models are optimized for deployment on a field-programmable gate array device. In this context, we demonstrate how latency and resource consumption scale with the input size and choice of algorithm. Moreover, the models proposed here are designed to work on the type of data and under the foreseen conditions at the CERN LHC during its high-luminosity phase. Through quantization-aware training and efficient synthetization for a specific field programmable gate array, we show that O(100)ns inference of complex architectures such as Deep Sets and Interaction Networks is feasible at a relatively low computational resource cost.  \nSubmitted to: Mach. Learn.: Sci. Technol.  \nUltrafast Jet Classification 2  \n1. Introduction  \nAt the CERN Large Hadron Collider (LHC), proton beams collide every 25 ns in each of the four particle detectors located around the LHC ring. The collision events generate sprays of outgoing particles that are detected by sensors, which amount to a data rate of tens of terabytes per second. For the ATLAS [1] and CMS [2] general-purpose experiments, the data throughput is too large to record every single event. Therefore, a subset of events are selected by a real-time event filtering system, called the trigger.  \nThe current trigger system consists of two stages. First, the Level-1 Trigger (L1T) reduces the event rate from O(10) MHz to O(100)kHz, rejecting ∼99.7% of all collisions. The frequency of collisions and limited buffer size set the maximum L1T latency to O(1) µs. Thus, the L1T is hardware based, with its algorithms running on Field-Programmable Gate Arrays (FPGAs) . The second stage is represented by the High-Level Trigger (HLT) . The HLT consists of software executed on a dedicated CPU farm and further reduces the event rate to 1 kHz. Only data accepted by the trigger system are saved entirely. Therefore, a high selection efficiency is of great importance for any LHC measurement and will become even more so after the high-luminosity upgrade.  \nThe LHC will undergo the High-Luminosity (HL-LHC) upgrade between 2026-2028 . The new HL-LHC will provide ten times more data. This will be achieved by increasing the number of simultaneous interactions per proton collision by a factor of three to four. To handle this upcoming increase in data complexity, the particle detectors at the LHC will be upgraded to maintain their detection efficiency for interesting physics processes. For the CMS experiment, this includes the addition of tracking information to the L1T, which will enable particle-level reconstruction and pileup mitigation as part of the L1T [3] . Consequently, Particle-Flow (PF) reconstruction [4] will be performed for the first time at the L1T, correlating tracks from the muon and tracking","cbCaih1FZnqmgpql","https://ap.wps.com/l/cbCaih1FZnqmgpql","pdf",1885897,1,17,"English","en",105,"# Introduction\n## Trigger system constraints at the LHC and HL-LHC\n## Need for jet origin identification\n## Deployment challenges for FPGA-based inference\n## Time multiplexing and resource limits","[{\"question\":\"What problem does this work address at the HL-LHC?\",\"answer\":\"It targets fast jet origin classification for use in the HL-LHC Level-1 trigger, where selection must be both accurate and extremely low-latency.\"},{\"question\":\"How are the machine learning models intended to be deployed?\",\"answer\":\"The models are optimized for execution on field-programmable gate array devices, with attention to how latency and resource consumption scale.\"},{\"question\":\"Which techniques enable ultrafast inference on FPGA hardware?\",\"answer\":\"The study uses quantization-aware training and efficient FPGA synthesis for specific hardware, making O(100) ns inference feasible for complex architectures like Deep Sets and Interaction Networks.\"}]","Ultrafast jet classification at the HL-LHC | PDF",1785681947,43,{"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},"ultrafast-jet-classification-at-the-hl-lhc","",{"@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/ultrafast-jet-classification-at-the-hl-lhc/118161/",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","2026-08-05","2026-08-02",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},"What problem does this work address at the HL-LHC?","Question",{"text":76,"@type":77},"It targets fast jet origin classification for use in the HL-LHC Level-1 trigger, where selection must be both accurate and extremely low-latency.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How are the machine learning models intended to be deployed?",{"text":81,"@type":77},"The models are optimized for execution on field-programmable gate array devices, with attention to how latency and resource consumption scale.",{"name":83,"@type":74,"acceptedAnswer":84},"Which techniques enable ultrafast inference on FPGA hardware?",{"text":85,"@type":77},"The study uses quantization-aware training and efficient FPGA synthesis for specific hardware, making O(100) ns inference feasible for complex architectures like Deep Sets and Interaction Networks.","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"]