Papers by Author | in ( Plenary | Parallel | Poster | Summary ) Sessions

Global VDET Pattern Recognition for ALEPH

             Andrew Bazarko(1), Dave Brown(2), Pascal Gay(3),
              Jean-Francois Pusztaszeri(1), Paul E. Rensing(1)

(1) European Laboratory for Particle Physics (CERN), 
        1211 Geneva 23, Switzerland

(2) Max-Planck-Institut f"ur Physik, Werner-Heisenberg-Institut,
            80805 M"unchen, Fed. Rep. of Germany,

(3) Laboratoire de Physique Corpusculaire, Universite Blaise Pascal,
         IN2P3-CNRS, Clermont-Ferrand, 63177 Aubiere, France

                        The ALEPH Collaboration
  • Paper (Postscript)
  • Paper (PDF)

  • Slides
                    
                                 Abstract
    
    With the current reliance on high-precision vertex detectors to provide very accurate information about the primary and secondary vertices in an event, the accuracy and efficiency of the assignment of vertex detector hits to tracks has become crucial.

    This paper discusses new software written for ALEPH which attempts to make this assignment in a global manner using linear programming techniques.

    The goal of the algorithm is to find the best assignment of hits to tracks, as measured by the sum of the track fit O2s and the number of hits assigned. A crucial step is the partitioning of the event into sets of tracks which are in competition for VDET hits and which have similar uncertainties at the VDET. The code then finds the best assignment of hits to tracks within each set. While a large number of problems can be solved by exhaustive search of all feasible patterns, about a quarter of the events contain a set of tracks whose solution is completely beyond the scope of this method.

    In such cases, the solution is found by a "branch and bound" algorithm interfaced with a state-of-the-art linear programming solver. No problem yet seen takes more than a few seconds of CPU time on a later generation RISC platform to solve.

    The performance of the new code will be compared to ALEPH's old, sequential pattern recognition algorithm using data and Monte Carlo. In addition, comparisons will be presented using Monte Carlo of ALEPH's new VDET and LEP II events.


    Submitter's Name: Paul E. Rensing
    Submitter's Institution: CERN
    Address of Institution: CERN Meyrin, CH-1211 Geneve 23, Switzerland
    Submitter's EMAIL address: rensing@cern.ch
    Submitter's telephone number: [41] 22 767 81 31
    FAX number: [41] 22 783 06 72