Clustering hits and predictions in data from TimePix3 detectors

• Název práce v češtině: Shlukování hitů a predikce v datech z detektoru TimePix3
• Název v anglickém jazyce: Clustering hits and predictions in data from TimePix3 detectors
• Klíčová slova: shlukování|TimePix3|experimentální fyzika|elementární částice|Strojové učení
• Klíčová slova anglicky: clustering|TimePix3|Experimental physics|elementary particles|Machine learning
• Akademický rok vypsání: 2021/2022
• Typ práce: diplomová práce
• Jazyk práce: angličtina
• Ústav: Katedra softwaru a výuky informatiky (32-KSVI)
• Vedoucí / školitel: RNDr. František Mráz, CSc.
• Řešitel: skrytý - zadáno a potvrzeno stud. odd.
• Datum přihlášení: 22.09.2022
• Datum zadání: 22.09.2022
• Datum potvrzení stud. oddělením: 06.10.2022
• Datum a čas obhajoby: 05.09.2023 09:00
• Datum odevzdání elektronické podoby: 20.07.2023
• Datum odevzdání tištěné podoby: 24.07.2023
• Datum proběhlé obhajoby: 05.09.2023
• Oponenti: RNDr. Tomáš Holan, Ph.D.
• Konzultanti: Benedikt Bergmann

Zásady pro vypracování

Electronic TimePix detectors can detect charged elementary particles with an array of pixels. During measurements, particles deposit charge on the pixels of a detector. A readout device processes the detected charges to generate a stream of so-called hits with information about the deposited charge. For further processing, we must cluster together all hits corresponding to an elementary particle with possible secondary particles when the particle decays. There are programs for clustering the hits detected by TimePix detectors. However, in many experiments, such clustering is much slower than the frequency of detected hits, effectively being the bottleneck of the whole processing.

The thesis aims to develop faster clustering algorithms for efficiently processing raw data from TimePix detectors. One of the possible approaches is selective clustering – trying to detect only clusters with a certain property like high energy while ignoring or only partially detecting clusters with small deposited energy. Another option is using the fact that high-energy pixels tend to be surrounded by other lower-energy pixels (often referred to as the halo effect) to simplify and possibly speed up the clustering algorithm. Furthermore, there is also a possibility of the parallelization of the clustering process.

When applying selective clustering, it is still desirable to know some properties like the composition of the whole stream of detected particles. The author should propose machine learning methods for predicting such properties based on the detected clusters and some features of the stream of hits. The solution includes proposing suitable easy-to-compute features and prediction methods.

The proposed algorithms will be implemented and compared with the existing clustering software to evaluate their efficiency and detection accuracy.

Seznam odborné literatury

B. Bergmann, M. Pichotka, S. Pospisil, J. Vycpalek, P. Burian, P. Broulim, J Jakubek: 3D track reconstruction capability of a silicon hybrid active pixel detector. The European Physical Journal C, Vol. 77, No. 6, 2017, pp. 421, accessible online https:/ /link.springer.com/article/10.1140/epjc/s10052-017-4993-4.

P. Burian, P. Broulím, M. Jára, V. Georgiev, B. Bergmann: Katherine: Ethernet Embedded Readout Interface for Timepix3. Journal of Instrumentation, Vol. 12, No. 11, 2017, pp. C11001, accessible online http:/ /stacks.iop.org/1748-0221/12/i=11/a=C11001.

P. Mánek: A system for 3D localization of gamma sources using Timepix3-based Compton cameras. Master thesis, Faculty of Mathematics and Physics, Charles University, 2018.

L. Meduna, B. Bergmann, P. Burian, P. Mánek, S. Pospíšil, M. Suk: Real-time Timepix3 data clustering, visualization and classification with a new Clusterer framework. Proceedings of Connecting the Dots and Workshop on Intelligent Trackers (CTD/WIT 2019), 2019, accessible online https:/ /doi.org/10.48550/arXiv.1910.13356.