1,721,471 research outputs found
2nd IML Machine Learning Workshop
No full textAt HL-LHC, the seven-fold increase of multiplicity wrt 2018 conditions poses a severe challenge to ATLAS and CMS tracking experiments. Both experiment are revamping their tracking detector, and are optimizing their software. But are there not new algorithms developed outside HEP which could be invoked: for example MCTS, LSTM, clustering, CNN, geometric deep learning and more?
We organize on the Kaggle platform a data science competition to stimulate both the ML and HEP communities to renew core tracking algorithms in preparation of the next generation of particle detectors at the LHC.
In a nutshell : one event has 100.000 3D points ; how to associate the points onto 10.000 unknown approximately helicoidal trajectories ? avoiding combinatorial explosion ? you have a few seconds. But we do give you 100.000 events to train on.
We ran ttbar+200 minimum bias event into ACTS a simplified (yet accurate) simulation of a generic LHC silicon detectors, and wrote out the reconstructed hits, with matching truth. We devised an accuracy metric which capture with one number the quality of an algorithm (high efficiency/low fake rate).
The challenge will run in two phases: the first on accuracy (no stringent limit on CPU time), starting in April 2018, and the second (starting in the summer 2018) on the throughput, for a similar accuracy
Mesure de la section éfficace de production de paires de quarks top/anti-top dans des collisions protons/anti-protons à \/s égale à 1.96 TeV auprès de l'expérience D0.
No full textThe top quark (t) discovered in 1995, could manifest the existence of new interactions. The ttbar pair production is a background in analysis that search for rare decays.The Tevatron is a proton/antu-proton collider with an energy of 1.96 TeV in the centre of mass. It produces ttbar paris with a theoritical cross section of 7 pb.The ttbar production cross section is measured in the "electron+jets" channel with the D0 detector, using an integrated luminosity of 360pb-1. The topological differences are utilised to estimate the number of signal events within a sample enriched in W boson electronic decays. The result is 9.0 +/- 2.0 +/- 1.7 +/- 0.- pb.The Calorimeter is essential in this measurement. The non-linearities and the gains of the calorimeter readout electronics are calibrated. The noise suppression algorigthm T42 is presented. It keeps cells with a signal above 4 sigma and their direct neighbors with a signal above 2.5sigma.Le quark top (t) découvert en 1995, pourrait être le sujet de la manifestation de nouvelles interactions et sa production en paires ttbar est un bruit de fond pour la recherche de processus rares.Le Tevatron, collisionneur proton/anti-proton avec une énergie dans le centre de masse de 1.96 TeV, produit des paires top/anti-top avec une section efficace théorique de 7 pb.La section efficace de production ttbar est mesurée dans le canal "électron+jets" en utilisant une luminosité intégrée de 360 pb-1 prise avec le détecteur D0. Dans un lot enrichi en désintégrations électronique de bosons W, le nombre d'évènments de signal est estimé grâce à des critères topologiques. La mesure est 9.0 +/- 2.0 +/- 1.7 +/- 0.- pb.La calibration des gains et des non-linéarités de l'électronique du calorimètre, essentiel à cette mesure, est effectuée. L'algorithme T42 est présenté, il permet une suppression du bruit électronique du calorimètre en conservant les signaux de plus de 4sigma et leurs voisins directs de plus de 2.5sigma
Introduction to Machine Learning
No full textLecture Content
Recent advancements in machine learning made it a major tool in big data analytics. We daily face applications of deep learning in image, speech and video recognitions, pledging for the efficiency of these methods in learning various complex tasks from data itself. With much care, advanced machine learning techniques are growingly used in Science, yielding better results than otherwise possible to Physicists. In this lecture, we will introduce the basics of machine learning and illustrate collider-specific aspects of deep learning by reviewing state-of-the-art applications of machine learning in high energy physics. The introduction to Machine Learning from Glen Cowan https://indico.cern.ch/event/1132551/ is highly recommended and repeated content will be avoided wherever possible.
Lecturer Bio
Vlimant is a research scientist in the Physics Mathematic and Astronomy department at the California Institute of Technology. Vlimant holds a Master in Quantum Mechanics from Ecole Normal Superieure of Paris and a Ph.D in particle physics from the Pierre et Marie Curie University. Vlimant is taking leading roles in the High Energy Physics community effort of developing deep learning and quantum computing applications for particle physics, actively involved with activities at CERN, the Compact Muon Solenoid collaboration and high performance computing facilities worldwide. Vlimant is the co-coordinator of the machine learning group in CMS. Vlimant is specialty section chief editor of “Big data and AI in High Energy Physics” in the frontiers in big data journal.
</p
Tracking at Hadron Colliders with Machine Learning
No full text
The reconstruction of charged particle trajectories is one of the main requirement for being able to achieve the research goals in collider physics. The resolution on kinematics obtained at low transverse momentum is crucial to many analysis, in particular in the calculation of transverse missing energy and identification of primary vertex. The canonical algorithm implemented in the experiment is based on a seeded combinatorial trajectory following using Kalman filter formalism to update trajectory parameters with sequence of measurements. The algorithm suffers, by construction, from combinatorial explosion and run-time is scaling worse than quadratically with the number concurrent collisions, tracks and hits. With the ever increasing performance of the LHC, and stagnation of computing funding, we are facing a tension between the computation needs and computing budget. While other ways of speeding up the algorithms are pursued ; part of the community is turning to machine learning and other pattern recognition technique to provide faster algorithms for track reconstruction with a view to reconciling costs and budgets. In this seminar I will review utilization of machine learning and advanced technique in tracking-like algorithms, underlying the challenges, promising solutions and possible future research on the topic.
</div
Voxxed Days CERN 2019
No full textCollisions at the CERN Large Hadron Collider (LHC) produce showers of particles that are detected by heterogenous detectors composed of hundreds of millions of individual sensors, laid out under complex geometry. An event can be seen as a tree of detectable particles branching from the unstable particles (e.g., the Higgs boson) produced in the collisions. Once detected, events are collected as arrays of isolated hits, which are then collectively processed to reconstruct the trajectory and energy of the particles that created them. In this contribution, we describe how the reconstruction and identification of these particles can be performed using graph networks. Given their capability of learning sparse representations, graph networks are ideal tools to create a fixed-geometry representation of an event, abstracting from the irregular geometry of the detectors used at colliders. As a first processing step of raw data, they provide an interface between particle detection and more rigid deep learning techniques, e.g., convolutional neural networks. In this respect, they represent a step forward to realistic deep learning applications for collider physics. As examples, we consider the task of reconstructing the trajectory of charged particles bending in the magnetic field of the detectors (tracking), denoising from parasitic collision (pile-up mitigation) and the identification of heavy particles (e.g., Higgs bosons) from the spray of particles that they produce (jets), where state-of-the-art performances are achieved
Data Science @ LHC 2015 Workshop
No full textNeuromorphic silicon chips have been developed over the last 30 years, inspired by the design of biological nervous systems and offering an alternative paradigm for computation, with real-time massively parallel operation and potentially large power savings with respect to conventional computing architectures. I will present the general principles with a brief investigation of the design choices that have been explored, and I'll discuss how such hardware has been applied to problems such as classification
Software and experience with managing workflows for the computing operation of the CMS experiment
No full textWe present a system deployed in the summer of 2015 for the automatic assignment of production and reprocessing workflows for simulation and detector data in the frame of the Computing Operation of the CMS experiment at the CERN LHC. Processing requests involves a number of steps in the daily operation, including transferring input datasets where relevant and monitoring them, assigning work to computing resources available on the CMS grid, and delivering the output to the Physics groups. Automation is critical above a certain number of requests to be handled, especially in the view of using more efficiently computing resources and reducing latency. An effort to automatize the necessary steps for production and reprocessing recently started and a new system to handle workflows has been developed. The state-machine system described consists in a set of modules whose key feature is the automatic placement of input datasets, balancing the load across multiple sites. By reducing the operation overhead, these agents enable the utilization of more than double the amount of resources with robust storage system. Additional functionality were added after months of successful operation to further balance the load on the computing system using remote read and additional resources. This system contributed to reducing the delivery time of datasets, a crucial aspect to the analysis of CMS data. We report on lessons learned from operation towards increased efficiency in using a largely heterogeneous distributed system of computing, storage and network elements
Data Preparation for the CMS detector at 8TeV at the LHC
No full textThe CMS detector, currently taking data at the LHC in Geneva, is a very complex apparatus composed of more than 70 million acquisition channels.
Fast and efficient methods for the calibration and the alignment of the detector are a key asset to exploit its full physics potential. Moreover, a reliable infrastructure for the monitoring of the data quality and for their validation are instrumental to ensure timely preparation of results for conferences and publications.
The CMS experiment has set up a powerful framework in order to cope with all these requirements and in 2012 it had to consolidate and optimize all the workflows to withstand the higher luminosity and energy delivered by the LHC machine.
The reconstruction algorithms have been optimized for the higher occupancies without compromising the physics performance. A MonteCarlo production with a statistic comparable to the collision data has been prepared and fully validated.
This contribution will cover the development and operational aspects of the offline workflows reporting about the CMS performance and the experience gained during the data taking
- …
