prikaz prve stranice dokumenta Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices
  Download
PDF 11.1 MB
doctoral thesis
Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices
2024. urn:nbn:hr:179:060086
Burazin Mišura, Arijana
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture

Cite this document

Burazin Mišura, A. (2024). Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices (Doctoral thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:060086

Burazin Mišura, Arijana. "Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices." Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2024. https://urn.nsk.hr/urn:nbn:hr:179:060086

Burazin Mišura, Arijana. "Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices." Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2024. https://urn.nsk.hr/urn:nbn:hr:179:060086

Burazin Mišura, A. (2024). 'Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices', Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 03 April 2025, https://urn.nsk.hr/urn:nbn:hr:179:060086

Burazin Mišura A. Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices [Doctoral thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2024 [cited 2025 April 03] Available at: https://urn.nsk.hr/urn:nbn:hr:179:060086

A. Burazin Mišura, "Artificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices", Doctoral thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2024. Available at: https://urn.nsk.hr/urn:nbn:hr:179:060086

Please login to the repository to save this object to your list.
Metadata
TitleArtificial Intelligence Algorithms for the Real-Time Selection of Physically Important Events Within the CMS L1 Trigger FPGA Devices
Title (english)Algoritmi umjetne inteligencije za selekciju fizikalno bitnih događaja u realnom vremenu na FPGA uređajima unutar CMS L1 okidača
AuthorArijana Burazin Mišura
MentorJosip Musić (mentor)
Committee memberVladan Papić (predsjednik povjerenstva)
Committee memberDamir Lelas (član povjerenstva)
Committee memberMaja Braović (član povjerenstva)
Committee memberJean-Baptiste Sauvan https://orcid.org/0000-0002-1086-9880 (član povjerenstva)
Committee memberMirela Kundid Vasić https://orcid.org/0000-0002-8441-0693 (član povjerenstva)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Split
Defense date and country2024, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Artificial Intelligence
Universal decimal classification
(UDC)		004 - Computer science and technology. Computing. Data processing
Abstract
The Large Hadron Collider is the largest particle accelerator in the world with the purpose of colliding protons accelerated to nearly the speed of light, resulting in the creation of new and exotic particles. Four detectors surround the collision points and record the data resulting from the proton-proton collisions. This data is then used for a wide range of physical analyses that intend to improve our understanding of particle physics and predict new theories. To increase its discovery potential, the LHC project aims to significantly boost the luminosity in the High Luminosity-Large Hadron Collider phase, which will, in turn, increase the number of collision events. The detector technologies are improved to address the challenges posed by the increased luminosity. However, it has become imperative to optimize the algorithms in the data selection system to effectively manage the presence of multiple interactions in the same bunch crossing (pile-up). This research focuses on using neural networks to select physically important events in real-time in the Compact Muon Solenoid, one of the LHC detectors. Simple multilayer perceptron and convolutional neural networks are developed to successfully distinguish electromagnetic showers from different types of backgrounds using raw detector images. Data selection and quantization methods are proposed to reduce the size of the presented models. Finally, profiling tools enable the determination of the precision settings needed to keep model inference accuracy, which further reduces the memory and computational requirements. The suggested approach was assessed using simulated data since the detector is still in the production phase. The inference time was less than 1 microsecond and signal-background classification achieved a classification accuracy of 97.01% for 2-bit-only quantization. This accuracy is the same as for the full-precision standard network, with a slightly decreased false negative rate. Overall, the research output presented within the thesis confirms the successful implementation of developed models in the targeted device while meeting the latency requirement, which makes them candidates for one of the level 1 trigger algorithms.
Abstract (english)
Veliki Hadronski Sudarač (engl. Large Hadron Collider, LHC) je najveći akcelerator čestica na svijetu. U njemu se sudaraju protoni, ubrzani skoro do brzine svjetlosti, što rezultira stvaranjem novih i egzotičnih čestica. Četiri detektora okružuju točke sudara i bilježe dobivene podatke iz proton-proton sudara. Ti se podaci zatim koriste za širok raspon fizikalnih analiza s ciljem poboljšanja našeg razumijevanja i predviđanja različitih teorija fizike čestica. Kako bi se povećala mogućnost otkrića, LHC projekt planira značajno povećati luminozitet u HL-LHC (engl. High Level LHC), što će rezultirati povećanjem broja sudara. Kako bi bilo moguće pratiti izazove koje predstavlja povećanje luminoziteta, trebalo je poboljšati tehnologije koje se koriste unutar detektora. Međutim, to ipak nije dovoljno, pa je imperativ postala optimizacija algoritama u sustavu odabira podataka kako bi se učinkovito upravljalo prisutnošću višestrukih interakcija u istom skupu sudara (pile-up). Ovo istraživanje usmjereno je na korištenje neuronskih mreža u odabiru fizički važnih događaja u stvarnom vremenu u Kompaktnom Muonskom Solenoidu, jednom od LHC detektora. Razvijene su jednostavne višeslojne perceptronske i konvolucijske neuronske mreže koje uspješno razlikuju elektromagnetske pljuskove od različitih vrsta pozadine koristeći neobrađene slike detektora. Predložene su metode odabira podataka i kvantizacije koje smanjuju veličinu predstavljenih modela. Konačno, alati za profiliranje omogućuju utvrđivanje postavki preciznosti potrebnih za održavanje točnosti zaključivanja modela, što dodatno smanjuje memorijske i računalne zahtjeve. Predloženi pristupi ispitani su korištenjem simuliranih podataka budući da je detektor još uvijek u fazi izrade. Vrijeme zaključivanja bilo je manje od 1 mikrosekunde te je kod klasifikacije signala i pozadine postignuta točnost od 97,01% pri kvantizaciji modela i podataka na samo 2 bita. Ova točnost je ista kao i za standardnu mrežu pune preciznosti, međutim s blagim padom stope lažno negativnih rezultata. Sveukupno, rezultati istraživanja predstavljeni u okviru teze potvrđuju uspješnu implementaciju razvijenih modela u ciljani uređaj uz ispunjavanje zahtjeva latencije, što ih čini kandidatima za jedan od algoritama okidanja razine 1.
Keywords
Keywords (english)
Languageenglish
URN:NBNurn:nbn:hr:179:060086
Promotion2024
Study programmeTitle: Electrical Engineering and Information Technology
Study programme type: university
Study level: postgraduate
Academic / professional title: doktor/doktorica znanosti, područje tehničkih znanosti, polje elektrotehnika (doktor/doktorica znanosti, područje tehničkih znanosti, polje elektrotehnika)
Type of resourceText
File originBorn digital
Access conditionsOpen access
Terms of useLicence In copyright icon
Created on2024-10-17 12:54:46