Eur. Phys. J. C (2015) 75:439
DOI 10.1140/epjc/s10052-015-3659-3
Regular Article - Experimental Physics
The Pandora software development kit for pattern recognition
J. S. Marshall
a
,M.A.Thomson
Cavendish Laboratory, University of Cambridge, Cambridge, UK
Received: 18 June 2015 / Accepted: 4 September 2015 / Published online: 21 September 2015
© The Author(s) 2015. This article is published with open access at Springerlink.com
Abstract The development of automated solutions to pat-
tern recognition problems is important in many areas of sci-
entific research and human endeavour. This paper describes
the implementation of the Pandora software development kit,
which aids the process of designing, implementing and run-
ning pattern recognition algorithms. The Pandora Applica-
tion Programming Interfaces ensure simple specification of
the building-blocks defining a pattern recognition problem.
The logic required to solve the problem is implemented in
algorithms. The algorithms request operations to create or
modify data structures and the operations are performed by
the Pandora framework. This design promotes an approach
using many decoupled algorithms, each addressing specific
topologies. Details of algorithms addressing two pattern
recognition problems in High Energy Physics are presented:
reconstruction of events at a high-energy e
+
e
−
linear col-
lider and reconstruction of cosmic ray or neutrino events in
a liquid argon time projection chamber.
1 Introduction
Pattern recognition is the identification of structures or regu-
larities in data. Problems requiring a pattern recognition solu-
tion occur in all areas of scientific research and our every-
day lives. This document describes the implementation of
the Pandora software development kit (SDK), which aims
to ease the process of designing, implementing and running
pattern recognition algorithms. The Pandora SDK was cre-
ated to address the problem of identifying energy deposits
from individual particles in fine granularity detectors in High
Energy Physics (HEP). The ideas described in this document
are, however, actually quite generic, covering a wide array
of problems where the aim is to sort points in time or space
into higher-level structures.
a
e-mail: marshall@hep.phy.cam.ac.uk
Figure 1 illustrates two typical pattern recognition prob-
lems in HEP. Figure 1a shows the simulated detector response
to the production and hadronic decay of Higgs and Z bosons
following high energy e
+
e
−
collisions at the Compact Lin-
ear Collider (CLIC). In order to extract measurements of
the Higgs boson properties, such as its coupling strengths,
it is vital to reconstruct and classify the individual particles
in large samples of events. Figure 1b shows the simulated
response of a Liquid Argon Time Projection Chamber (LAr
TPC) to a charged current electron neutrino interaction. In
order to understand neutrino mixing and CP-violation in the
neutrino sector, it is crucial to identify and characterise each
particle in this challenging topology.
The idea underpinning the Pandora SDK is that the inter-
faces for pattern recognition problems are well defined, as are
the operations that must be performed by pattern recognition
algorithms. Whoever poses the pattern recognition problem
must specify the building-blocks, or space-points, that define
the problem. They must also be able to extract the output
structures, such as clusters, that represent the solution. The
algorithms that address the problem must be able to build
clusters of space-points and should be able to manipulate
clusters by splitting them up or merging them together. What
differs between pattern recognition problems is the precise
logic controlling the algorithm operations.
2 Historical context
The Pandora project began in 2007 to provide the first parti-
cle flow calorimetry implementation for the proposed Inter-
national Linear Collider (ILC). A particle flow algorithm
was developed, exploiting the fine granularity detectors in
order to reconstruct the paths of individual visible particles.
Successful identification of the trajectories allows particle
four-momenta to be extracted from the subdetector system
in which they are best-measured, delivering unprecedented
jet energy resolution. The Pandora algorithm was used to
123
剩余15页未读，继续阅读
资源评论
