Physics Data Analysis Research Articles

A Comparison of Data-Access Platforms for the Computing of Large Hadron Collider Experiments

Performance, reliability and scalability in data-access are key issues in the context of the computing Grid and High Energy Physics data processing and analysis applications, in particular considering the large data size and I/O load that a Large Hadron Collider data centre has to support. In this paper we present the technical details and the results of a large scale validation and performance measurement employing different data-access platforms-namely CASTOR, dCache, GPFS and Scalla/Xrootd. The tests have been performed at the CNAF Tier-1, the central computing facility of the Italian National Institute for Nuclear Research (INFN). Our storage back-end was based on Fibre Channel disk-servers organized in a Storage Area Network, being the disk-servers connected to the computing farm via Gigabit LAN. We used 24 disk-servers, 260 TB of raw-disk space and 280 worker nodes as computing clients, able to run concurrently up to about 1100 jobs. The aim of the test was to perform sequential and random read/write accesses to the data, as well as more realistic access patterns, in order to evaluate efficiency, availability, robustness and performance of the various data-access solutions.

A software Event Summation System for MDSplus

The MDSplus data acquisition and management system uses software events for communication among interdependent processes anywhere on the network. Actions can then be triggered, such as a data-acquisition routine, or analysis or display programs waiting for data. A small amount of data, such as a shot number, can be passed with these events. Since programs sometimes need more than one data set, we developed a system on NSTX to declare composite events using logical AND and OR operations. The system is written in the IDL language, so it can be run on Linux, Macintosh or Windows platforms. Like MDSplus, the Experimental Physics and Industrial Control System (EPICS) is a core component of the NSTX software environment. The Event Summation System provides an IDL-based interface to EPICS. This permits EPICS-aware processes to be synchronized with MDSplus-aware processes, to provide, for example, engineering operators information about physics data acquisition and analysis. Reliability was a more important design consideration than performance for this system; the system's architecture includes features to support this. The system has run for weeks at a time without requiring manual intervention. Hundreds of incoming events per second can be handled reliably. All incoming and declared events are logged with a timestamp. The system can be configured easily through a single, easy-to-read text file.

High Energy Physics event selection with Gene Expression Programming

Gene Expression Programming is a new evolutionary algorithm that overcomes many limitations of the more established Genetic Algorithms and Genetic Programming. Its application to event selection in high energy physics data analysis is presented using as an example application the selection of K S particles produced in e + e − interactions at 10 GeV and reconstructed in the decay mode K S → π + π − . The algorithm was used for automatic identification of classification criteria for signal/background separation. For the problem studied and for data samples with signal to background ratios between 0.25 and 5, the classification accuracy obtained with the criteria developed by the GEP algorithm was in the range of 92–95%.

Implementation of the SVM algorithm for high energy physics data analysis

Elementary particle physics experiments, searching for very rare processes require the efficientanalysis and selection algorithms able to separate signal from the overwhelming background. Inthe last ten years a number of powerful kernel-based learning machines, like Support VectorMachines (SVM), have been developed. SVM approach to signal and background separation isbased on building a separating hyperplane defined by the support vectors. The margin betweenthem and the hyperplane is maximized. The extensions to a non-linear separation are performed bymapping the input vectors into a high dimensional space, in which data can be linearly separated.The use of kernel functions allows us to perform computations in a high dimension feature spacewithout explicitly knowing a mapping function.We have implemented an SVM algorithm and integrated it with the CERN ROOT package,which is currently a standard analysis tool used by elementary particle physicists. We also used theimplemented SVM package to identify hadronic decays of τ leptons in the ATLAS experiment atLHC accelerator. The performance of the method is compared to the likelihood estimator, whichdoes not take into account correlations between variables. The use of SVM significantly reducesthe number of background events.

New Developments of the Goodness-of-Fit Statistical Toolkit

The Statistical Toolkit is a project for the development of open source software tools for statistical data analysis in experimental particle and nuclear physics. The second development cycle encompassed an extension of the software functionality and new tools to facilitate its usage in experimental environments. The new developments include additional goodness-of-fit tests, new implementations of existing tests to improve their statistical precision or computational performance, a new component to extend the usability of the toolkit with other data analysis systems, and new tools for an easier configuration and build of the system in the user's computing environment. The computational performance of all the algorithms implemented has been studied

Finnish CMS-TOB cosmic rack

We present a cosmic rack, the FinnCRack. This device is a silicon strip detector-based telescope that measures tracks of cosmic particles. The FinnCRack is constructed using components of the Tracker Outer Barrel (TOB) of the CMS experiment at the CERN LHC. The device is part of the TOB integration and verification effort together with its sister telescope, the CERN CRack. Both CRacks mimick a six degree slice of the TOB barrel structure. The FinnCRack is intended to (a) serve as a platform for TOB software development, both analysis and online software such as run control; (b) be used for noise and cluster shape studies; (c) act as a reference tracker in detector studies; and (d) provide a testbed for track-based alignment testing and development. The construction and setup of the FinnCRack have been documented in detail—the entire chain from connecting cables to physics data analysis—and the operation guide was tested in practice. Both these actions serve the purpose of training and attracting future HEP students. We also showed that we were able to measure cosmic muon tracks.

Gene Expression Programming Approach to Event Selection in High Energy Physics

Gene Expression Programming is a new evolutionary algorithm that overcomes many limitations of the more established Genetic Algorithms and Genetic Programming. Its first application to high energy physics data analysis is presented. The algorithm was successfully used for event selection on samples with both low and high background level. It allowed automatic identification of selection rules that can be interpreted as cuts applied on the input variables. The signal/background classification accuracy was over 90% in all cases

The laser calibration system of the ALICE time projection chamber

A Large Ion Collider Experiment (ALICE) is the only experiment at the Large Hadron Collider (LHC) dedicated to the study of heavy ion collisions. The Time Projection Chamber (TPC) is the main tracking detector covering the pseudo rapidity range ¦η¦ < 0.9. It is designed for a maximum multiplicity dN/dy = 8000. The aim of the laser system is to simulate ionizing tracks at predifined positions throughout the drift volume in order to monitor the TPC response to a known source. In particular, the alignment of the read-out chambers will be performed, and variations of the drift velocity due to drift field imperfections can be measured and used as calibration data in the physics data analysis. In this paper we present the design of the pulsed UV laser and optical system, together with the control and monitoring systems.

The LCG PI project: using interfaces for physics data analysis

In the context of the LHC computing grid (LCG) project, the applications area develops and maintains that part of the physics applications software and associated infrastructure that is shared among the LHC experiments. The "physicist interface" (PI) project of the LCG application area encompasses the interfaces and tools by which physicists will directly use the software, providing implementations based on agreed standards like the analysis systems subsystem (AIDA) interfaces for data analysis. In collaboration with users from the experiments, work has started with implementing the AIDA interfaces for (binned and unbinned) histogramming, fitting and minimization as well as manipulation of tuples. These implementations have been developed by re-using existing packages either directly or by using a (thin) layer of wrappers. In addition, bindings of these interfaces to the Python interpreted language have been done using the dictionary subsystem of the LCG applications area/SEAL project. The actual status and the future planning of the project will be presented

Search for [formula omitted] and [formula omitted] using genetic programming event selection

We apply a genetic programming technique to search for the doubly Cabibbo suppressed decays Λc+→pK+π− and Ds+→K+K+π−. We normalize these decays to their Cabibbo favored partners and find BR(Λc+→pK+π−)/BR(Λc+→pK−π+)=(0.05±0.26±0.02)% and BR(Ds+→K+K+π−)/BR(Ds+→K−K+π+)=(0.52±0.17±0.11)% where the first errors are statistical and the second are systematic. Expressed as 90% confidence levels (CL), we find <0.46 and <0.78%, respectively. This is the first successful use of genetic programming in a high energy physics data analysis.

Trends in physics data analysis algorithms

The paper provides a new look at algorithmic trends in modern physics experiments. Based on recently presented material, it attempts to draw conclusions in order to form a coherent historical picture of the past, present, and possible future of the field of data analysis techniques in physics. The importance of cross disciplinary approaches is stressed.

A multi-variate discrimination technique based on range-searching

We present a fast and transparent multi-variate event classification technique, called PDE-RS, which is based on sampling the signal and background densities in a multi-dimensional phase space using range-searching. The employed algorithm is presented in detail and its behaviour is studied with simple toy examples representing basic patterns of problems often encountered in High Energy Physics data analyses. In addition an example relevant for the search for instanton-induced processes in deep-inelastic scattering at HERA is discussed. For all studied examples, the new presented method performs as good as artificial Neural Networks and has furthermore the advantage to need less computation time. This allows to carefully select the best combination of observables which optimally separate the signal and background and for which the simulations describe the data best. Moreover, the systematic and statistical uncertainties can be easily evaluated. The method is therefore a powerful tool to find a small number of signal events in the large data samples expected at future particle colliders.

Redundant arrays of IDE drives

The next generation of high-energy physics experiments is expected to gather prodigious amounts of data. New methods must be developed to handle this data and make analysis at universities possible. We examine some techniques that use recent developments in commodity hardware. We test redundant arrays of integrated drive electronics (IDE) disk drives for use in offline high-energy physics data analysis. IDE redundant array of inexpensive disks (RAID) arrays prices now equal the cost per terabyte of million dollar tape robots! The arrays can be scaled to sizes affordable to institutions without robots and used when fast random access at low cost is important. We also explore three methods of moving data between sites; internet transfers, hot pluggable IDE disks in FireWire cases, and writable digital video disks (DVD-R) disks.

An ODMG-compatible testbed architecture for scalable management and analysis of physics data

This paper describes a testbed architecture for the investigation and development of scalable approaches to the management and analysis of massive amounts of high energy physics data. The architecture has two components: an interface layer that is compliant with a substantial subset of the ODMG-93 Version 1.2 specification, and a lightweight object persistence manager that provides flexible storage and retrieval services on a variety of single- and multi-level storage architectures, and on a range of parallel and distributed computing platforms.

Data analysis in relativistic nuclear physics using neural network

Two neural network algorithms for data analysis in relativistic nuclear physics are presented. A neural network technique (Hopfield method) is used in order to reconstruct particle tracks starting from a data set obtained with a coordinate detector system. An algorithm for circles recognition using deformable templates is carried out and its performances are studied. The technical limitations of the detectors, which in real situation prevent the possibility to reconstruct hits right on the circle, and presence of the noise points are taken into account.

Aerosol number size distributions from 3 to 500 nm diameter in the arctic marine boundary layer during summer and autumn

Aerosol physics measurements made onboard the Swedish icebreaker Oden in the late Summer and early Autumn of 1991 during the International Arctic Ocean Expedition (IAOE-91) have provided the first data on the size distribution of particles in the Arctic marine boundary layer (MBL) that cover both the number and mass modes of the size range from 3 to 500 nm diameter. These measurements were made in conjunction with atmospheric gas and condensed phase chemistry measurements in an effort to understand a part of the ocean-atmosphere sulphur cycle. Analysis of the particle physics data showed that there were three distinct number modes in the submicrometric aerosol in the Arctic MBL. These modes had geometric mean diameters of around 170 nm, 45 nm and 14 nm referred to as accumulation, Aitken and ultrafine modes, respectively. There were clear minima in number concentrations between the modes that appeared at 20–30 nm and at 80–100 nm. The total number concentration was most frequently between 30 and 60 particles cm-3 with a mean value of around 100 particles cm-3, but the hourly average concentration varied over two to three orders of magnitude during the 70 days of the expedition. On average, the highest concentration was in the accumulation mode that contained about 45% of the total number, while the Aitken mode contained about 40%. The greatest variability was in the ultrafine mode concentration which is indicative of active, nearby sources (nucleation from the gas phase) and sinks; the Aitken and accumulation mode concentrations were much less variable. The ultrafine mode was observed about two thirds of the time and was dominant 10% of the time. A detailed description and statistical analysis of the modal aerosol parameters is presented here.

Why isn’t every physicist a Bayesian?

Physicists embarking on seemingly routine error analyses are finding themselves grappling with major conceptual issues which have divided the statistics community for years. While the philosophical aspects of the debate may be endless, a practicing experimenter must choose a way to report results. The results can depend on which of the two major frameworks, classical or Bayesian, one adopts. This article reviews reasons why most data analysis in particle physics has traditionally been carried out within the classical framework, and why this will probably continue to be the case. However, Bayesian reasoning has recently made significant inroads in some published work in this field, and many other particle physicists may frequently think in a Bayesian manner without realizing it. I illustrate the issues involved with a few simple, commonly encountered examples which reveal how each framework can sometimes lead to unsatisfying results.

Multiquadric radial basis functions for representing multidimensional high energy physics data

We examine how basis functions can be used to represent multidimensional data and argue that radial basis functions are a natural choice. Of all the possible radial basis functions, multiquadrics, counter-intuitively, turn out to be supremely suitable for representing high energy physics event densities, and probably many other forms of scientific data, because they fulfill the physicists' requirements of sensible non-erratic behaviour and follow significant variations in a smooth and faithful way. The stochastic nature of high energy physics data often gives rise to unwelcome statistical fluctuations; our algorithm can be used to smooth these out. The smoothed function is much easier to visualise than the original scattered events. A suite of Fortran subroutines has been written in the framework of current high energy physics data analysis systems, HBOOK and PAW.

A derived-data type for data analysis

A frequent requirement in physics data analysis is the ability to form and manipulate histograms; this is done currently using calls to the HBOOK subroutine package. Could its interfaces and internal structure be improved by defining a histogram to be a Fortran 90 data type instead?

A CAMAC-VME-Macintosh data acquisition system for nuclear experiments

A multiprocessor system for data acquisition and analysis in low-energy nuclear physics has been realized. The system is built around CAMAC, the VMEbus, and the Macintosh PC. Multiprocessor software has been developed, using RTF, MACsys, and CERN cross-software. The execution of several programs that run on several VME CPUs and on an external PC is coordinated by a mailbox protocol. No operating system is used on the VME CPUs. The hardware, software, and system performance are described. >