Strojové učení obrázkových jazyků s aplikací v detekci elementárních částic

• Thesis title in Czech: Strojové učení obrázkových jazyků s aplikací v detekci elementárních částic
• Thesis title in English: Machine learning of picture languages with application in elementary particle detection
• Key words: strojové učení|obrázkové jazyky|elementární částice
• English key words: machine learning|picture languages|elementary particles
• Academic year of topic announcement: 2022/2023
• Thesis type: dissertation
• Thesis language: čeština
• Department: Department of Software and Computer Science Education (32-KSVI)
• Supervisor: RNDr. František Mráz, CSc.
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 15.09.2023
• Date of assignment: 15.09.2023
• Confirmed by Study dept. on: 21.09.2023
• Advisors: Benedikt Bergmann

Guidelines

Picture languages [2] generalize (one-dimensional) string languages into two dimensions. The strict formal definition of picture languages enables designing various types of automata for accepting them, like four-way finite automata, sgraffito automata [7], two-dimensional restarting automata [4], online tesselation automata [1], etc.

The thesis aims to propose new models of two-dimensional automata and picture languages and design methods for their automated construction using machine learning methods, e.g., by inferring such models from labeled examples [3]. The goal is to efficiently recognize suitable picture language classes applicable to theoretical and practical problems.
In particular, the developed model should enable inferring models for recognizing pictures of traces of elementary particles.

Recently developed detectors of elementary particles [5] enable to follow traces of hundreds of thousands of elementary particles per second with a time resolution of nanoseconds [6]. The traces of charged particles of different energies can be considered as pictures. With additional information, e.g., the amount of energy deposited in the detector, such pictures carry essential information about the detected particles.

Employing methods from machine learning could improve the throughput of systems for analyzing streams of particles by, e.g., effectively identifying particle classes or filtering unnecessary data. The student should review known approaches for particle detection using pixel detectors, analyze known algorithms for particle classification, and propose and implement new algorithms based on picture languages and picture automata to analyze existing and future data from particle detectors.

References

[1] B. Bergmann, P. Broulím, P. Burian, T. Čelko, P. Mánek, K. Gunthoti, F. Mráz, S. Pospíšil, M. Sitarz, P. Smolyanskiy, C. Søndergaard, S. Wender: A two-layer Timepix3 stack for improved charged particle tracking and radiation field decomposition

[2] D. Giammarresi and A. Restivo: Two-Dimensional Languages. In: Handbook of Formal Languages. Ed. by A. Salomaa and G. Rozenberg. Vol. 3 – Beyond Words. Springer-Verlag, 1997. Chap. 4, pp. 215–267.

[3] C. de la Higuera: Grammatical Inference: Learning Automata and Grammars. Cambridge University Press, 2010.

[4] L. Krtek and F. Mráz: Two-Dimensional Limited Context Restarting Automata. Fundamenta Informaticae 148.3–4 (2016), pp. 309–340.

[5] X. Llopart, R. Ballabriga, M. Campbell, L. Tlustos, W. Wong: Timepix, a 65k programmable pixel readout chip for arrival time, energy and/or photon counting measurement. In: Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 581.1 (2007). VCI 2007, pp. 485–494.

[6] 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. In: Proceedings of the CTD/WIT 2019, No. PROC-CTD19-105, Valencia, Spain April 2-5, arXiv 910.13356, 2019.