131 research outputs found
A muon telescope as demonstrator of the juno top tracker detector
No full textA four-layer muon telescope has been built, employing the equipment and electronics developed for the Top Tracker (TT) detector of the Jiangmen Underground Neutrino Observatory (JUNO). It will serve as a demonstrator of the hardware capabilities in terms of detection efficiency, processing power and system reliability. The entire readout, trigger and acquisition systems have been conceived and built around versatile modular electronics embedding the latest generation of system on chips. A detailed description of the telescope will be given along with the status of the TT electronics and their preliminary validation tests
Metalearning: a survey of trends and technologies
No full textMetalearning attracted considerable interest in the machine learning community in the last years. Yet, some disagreement remains on what does or what does not constitute a metalearning problem and in which contexts the term is used in. This survey aims at giving an all-encompassing overview of the research directions pursued under the umbrella of metalearning, reconciling different definitions given in scientific literature, listing the choices involved when designing a metalearning system and identifying some of the future research challenges in this domain. © 2013 The Author(s)
An analytical model for photomultiplier tube calibration
No full textThe purpose of the present article is to offer an analytical model that can
be used for the calibration of photomultiplier tubes (PMTs). The derivation of
the mathematical formulae is discussed extensively and we apply this machinery
to a real problem; the gain determination of the R7081 Hamamatsu PMT. We hope
that our work can provide a third, powerful alternative to the task of gain
deconvolution. We also believe that the techniques explained in this article
can be applied to other problems in PMT calibration and monitoring
Calibration du détecteur de Double Chooz et étude du bruit de fond induit par un rayonnement cosmique
No full textDouble Chooz est une expérience de type court ligne de base , auprès des réacteurs nucléaires de Chooz, qui a été conçue pour observer les oscillations des neutrinos associées à l angle de mélange 13. Une partie de ma recherche était axée sur le développement du logiciel requis pour l'étalonnage de l Inner Veto de Double Chooz et à l'analyse des données qui y sont associées. J'ai été responsable des tests de qualité effectués sur tous les photomultiplicateurs avant leur installation. J'ai fait toutes les mesures nécessaires et j'ai analysé la majorité des données. Grâce à ce travail, j'ai extrait les premières valeurs des gains et j ai déterminé les valeurs nominales des hautes tensions nécessaires pour les photomultiplicateurs. Toutes ces données nous ont aidé dans la mise au point du détecteur. Aussi j'ai été le responsable de l'analyse de l extraction et stabilité des gains pendant l expérience. J'ai travaillé très activement sur l'analyse des données et en particulier sur l'évaluation des différentes sources de bruit de fond. J'ai développé plusieurs techniques pour isoler et étudier les muons cosmiques qui activent le détecteur Double Chooz. J'ai également travaillé sur l'estimation du taux de neutrons rapides. Les techniques que j'ai présentées ont été utilisées dans la première publication de Double Chooz. Finalement, j'ai étudié un bruit de fond instrumental important pour le détecteur Double Chooz. J'ai développé un nouvel ensemble d'algorithmes pour identifier et rejeter ces événements pathologiques utilisant l'orientation géométrique de la charge à l'intérieur du détecteur.Double Chooz is a short-baseline experiment, located at the Chooz power plant, designed to observe the neutrino oscillation signal controlled by the 13 mixing angle. Part of my scientific research, as a graduate student, was directed towards the development of the software needed for the calibration of the Double Chooz Inner Veto and the analysis of the data associated with this task. I was responsible for the quality tests performed in every photomultiplier prior to its installation. I completed all the necessary measurements and analysed the data, extracting the first set of gains and determining the nominal high voltage values needed to be applied in all photomultipliers. All this information served as valuable input to the detector configuration. I was also responsible for the Inner Veto photomultiplier gain analysis during the first months of data taking. I was also very actively involved in data analysis and the estimations of the various sources of background. I initiated a number of methods to isolate and study the cosmic muon events that activate the detector. Additionally I worked on the estimation of the fast neutron rate registered in the detector. The techniques I put forward played a key role and were used in the first Double Chooz publication. Finally, I developed a set of algorithms to identify and reject an instrumental background, relevant for the Double Chooz detector using topological information of the deposited charge.STRASBOURG-Bib.electronique 063 (674829902) / SudocSudocFranceF
Calibration of photomultiplier tubes
No full textThe purpose of the present article is to demonstrate the calibration of
photomultipliers with a gaussian single photoelectron response using a
numerical method based on the Discrete Fourier Transform (DFT). Conventional
techniques, commonly employed in the literature, use approximate models or
brute force numerical calculations of the convolution integrals that lead to
the charge response function of the photomultiplier, . In this
publication, we explain how a truncated gaussian model for the single
photoelectron amplification can lead to rigorous results if one leans on the
DFT approach. The distinct feature of this procedure is that is
calculated to all orders in the Poisson mean that characterizes the light
intensity and no approximations are needed. This scheme was applied to the
calibration of the Hamamatsu R7081 photomultiplier tube and a comparison of the
DFT approach with the more standard numerical integration method is also
presented. Last, toy Monte Carlo data were analyzed for different values of
to understand the precision of the DFT method
Querying and Merging Heterogeneous Data by Approximate Joins on Higher-Order Terms
No full textIntegrating heterogeneous data from sources as diverse as web pages, digital libraries, knowledge bases, the Semantic Web and databases is an open problem. The ultimate aim of our work is to be able to query such heterogeneous data sources as if their data were conveniently held in a single relational database. Pursuant to this aim, we propose a generalisation of joins from the relational database model to enable joins on arbitrarily complex structured data in a higher-order representation. By incorporating kernels and distances for structured data, we further extend this model to support approximate joins of heterogeneous data. We demonstrate the flexibility of our approach in the publications domain by evaluating example approximate queries on the CORA data sets, joining on types ranging from sets of co-authors through to entire publications.Integrating heterogeneous data from sources as diverse as web pages, digital libraries, knowledge bases, the Semantic Web and databases is an open problem. The ultimate aim of our work is to be able to query such heterogeneous data sources as if their data were conveniently held in a single relational database. Pursuant to this aim, we propose a generalisation of joins from the relational database model to enable joins on arbitrarily complex structured data in a higher-order representation. By incorporating kernels and distances for structured data, we further extend this model to support approximate joins of heterogeneous data. We demonstrate the flexibility of our approach in the publications domain by evaluating example approximate queries on the CORA data sets, joining on types ranging from sets of co-authors through to entire publications
Multi-task learning and transfer: The effect of algorithm representation
No full textExploring multiple classes of learning algorithms for those algorithms which perform best in multiple tasks is a complex problem of multiple-criteria optimisation. We use a genetic algorithm to locate sets of models which are not outperformed on all of the tasks. The genetic algorithm develops a population of multiple types of learning algorithms, with competition between individuals of different types. We find that inherent differences in the convergence time and performance levels of the different algorithms leads to misleading population effects. We explore the role that the algorithm representation and initial population has on task performance. Our findings suggest that separating the representation of different algorithms is beneficial in enhancing performance. Also, initial seeding is required to avoid premature convergence to non-optimal classes of algorithms
Neutrino mass ordering determination through combined analysis with JUNO and KM3NeT/ORCA
No full textInternational audienceThe determination of neutrino mass ordering (NMO) is one of the prime goals of several neutrinoexperiments.KM3NeT/ORCA and JUNO are two next-generation neutrino oscillation experiments both aiming ataddressing this question.ORCA determines the NMO by probing Earth matter effects on the oscillation ofatmospheric neutrinos in the GeV energy range.JUNO, on the other hand, is sensitive to the NMO by investigating the interference effects offast oscillations in the reactor electron antineutrino spectrum at medium baseline.This poster presents the potential of determining the NMO through a combined analysis ofJUNO and ORCA data.When measuring the with a wrong ordering assumption, the best-fit valuesare different between the two experiments.This tension, together with good constraints on the measurement by bothexperiments, enhances the combined NMO sensitivity beyond the simple sum of their sensitivities.The analysis shows that 5 significance is reachable in less than 2 years of data takingwith both experiements for true normal neutrino mass ordering assuming current global best-fitvalues of the oscillation parameters, while 6 years will be needed for any other parameter set
Pairwise meta-rules for better meta-learning-based algorithm ranking
No full textIn this paper, we present a novel meta-feature generation method in the context of meta-learning, which is based on rules that compare the performance of individual base learners in a one-against-one manner. In addition to these new meta-features, we also introduce a new meta-learner called Approximate Ranking Tree Forests (ART Forests) that performs very competitively when compared with several state-of-the-art meta-learners. Our experimental results are based on a large collection of datasets and show that the proposed new techniques can improve the overall performance of meta-learning for algorithm ranking significantly. A key point in our approach is that each performance figure of any base learner for any specific dataset is generated by optimising the parameters of the base learner separately for each dataset
- …
