Machine learning "as a service" for high energy physics (MLaaS4HEP): evolution of a framework for ML-based physics analyses

The scientific success of the LHC experiments at CERN highly depends on the availability of computing resources which efficiently store, process, and analyse the amount of data collected every year. This is ensured by the Worldwide LHC Computing Grid infrastructure that connect computing centres distributed all over the world with high performance network. LHC has an ambitious experimental program for the coming years, which includes large investments and improvements both for the hardware of the detectors and for the software and computing systems, in order to deal with the huge increase in the event rate expected from the High Luminosity LHC (HL-LHC) phase and consequently with the huge amount of data that will be produced. Since few years the role of Artificial Intelligence has become relevant in the High Energy Physics (HEP) world. Machine Learning (ML) and Deep Learning algorithms have been successfully used in many areas of HEP, like online and offline reconstruction programs, detector simulation, object reconstruction, identification, Monte Carlo generation, and surely they will be crucial in the HL-LHC phase. This thesis aims at contributing to a CMS R&D project, regarding a ML "as a Service" solution for HEP needs (MLaaS4HEP). It consists in a data-service able to perform an entire ML pipeline (in terms of reading data, processing data, training ML models, serving predictions) in a completely model-agnostic fashion, directly using ROOT files of arbitrary size from local or distributed data sources. This framework has been updated adding new features in the data preprocessing phase, allowing more flexibility to the user. Since the MLaaS4HEP framework is experiment agnostic, the ATLAS Higgs Boson ML challenge has been chosen as physics use case, with the aim to test MLaaS4HEP and the contribution done with this work.

Monte Carlo simulation of the WENDI-2 neutron dosimeter

Il presente lavoro di tesi, sviluppato nell’arco di sei mesi presso l’Institut Supérieur Industriel de Bruxelles (ISIB) in collaborazione con Ion Beam Application Group (IBA, Louvain la Neuve), ha come principale soggetto lo studio della risposta del rem meter WENDI-2 commercializzato da Thermo Scientific. Lo studio si è basato principalmente sull’uso del codice Monte Carlo MCNPX 2.5.0, simulando la risposta del detector sia in caso di campi di radiazione neutronica monoenergetici sia in corrispondenza di spettri neutronici continui. La prima fase è stata dedicata alla modellizzazione MCNPX del rem counter, consentendo così la valutazione della sua funzione risposta. Questa è stata ricostruita interpolando 93 punti, ciascuno calcolato in corrispondenza di un singolo valore di energia di una sorgente puntiforme, compreso tra 1 meV e 5 GeV. In tal caso è stata rilevata un’ottima corrispondenza tra i risultati ottenuti e quelli riportati nella letteratura scientifica esistente. In una seconda fase, al fine di ottenere informazioni sulla risposta di WENDI II in corrispondenza di campi complessi di radiazione, simulazioni MCNPX sono state realizzate riproducendo un ambiente di lavoro esistente presso la sede IBA di Louvain la Neuve: la risposta del detector è stata valutata in corrispondenza di 9 diverse posizioni all’interno di un bunker contenente un ciclotrone PET (18 MeV H-), implicando la rilevazione di campi di radiazione neutronica continui ed estesi dalle energie termiche fino a 18 MeV. I risultati ottenuti sono stati infine comparati con i valori di dose ambiente equivalente calcolata nelle stesse condizioni di irraggiamento.

Proposal of a Continuous Read-out Implementation in the ALICE-TOF detector

The main purpose of ultrarelativistic heavy-ion collisions is the investigation of the QGP. The ALICE experiment situated at the CERN has been specifically designed to study heavy-ion collisions for centre-of-mass energies up to 5.5 per nucleon pair. Extended particle identification capability is one of the main characteristics of the ALICE experiment. In the intermediate momentum region (up to 2.5 GeV/c for pi/K and 4 GeV/c for K/p), charged particles are identified in the ALICE experiment by the Time of Flight (TOF) detector. The ALICE-TOF system is a large-area detector based on the use of Multi-gap Resistive Plate Chamber (MRPC) built with high efficiency, fast response and intrinsic time resolution better than 40 ps. This thesis work, developed with the ALICE-TOF Bologna group, is part of the efforts carried out to adapt the read-out of the detector to the new requirements after the LHC Long Shutdown 2. Tests on the feasibility of a new read-out scheme for the TOF detector have been performed. In fact, the achievement of a continuous read-out also for the TOF detector would not be affordable if one considers the replacement of the TRM cards both for hardware and budget reasons. Actually, the read-out of the TOF is limited at 250 kHz i.e. it would be able to collect up to just a fourth of the maximum collision rate potentially achievable for pp interactions. In this Master’s degree thesis work, I discuss a different read-out system for the ALICE-TOF detector that allows to register all the hits at the interaction rate of 1 MHz foreseen for pp interactions after the 2020, by using the electronics currently available. Such solution would allow the ALICE-TOF detector to collect all the hits generated by pp collisions at 1 MHz interaction rate, which corresponds to an amount four times larger than that initially expected at such frequencies with the triggered read-out system operated at 250 kHz for LHC Run 3. The obtained results confirm that the proposed read-out scheme is a viable option for the ALICE TOF detector. The results also highlighted that it will be advantageous if the ALICE-TOF group also implement an online monitoring system of noisy channels to allow their deactivation in real time.