Run II Advanced Analysis Working Groups

The Advanced Analysis Group (Run2AAG), which  operates under the aegis of the Run II Advanced Analysis Workshop, is comprised of the following working groups

The Tools group will provide analysis interfaces to useful software; the Simulation group will make available Monte Carlo simulations ofhigh energy events; the Evaluation group will assess different multivariate methods by applying them to physics topics of current interest and the Statistics group will address statistical issues arisingfrom these studies and others.  We briefly describe belowthe scope of each working group. For more detailed information please consult the relevant group's web pages.


 

Tools

Members

Pushpa Bhat
Catalin Ciobanu
Phil Koehn
Jim Kowalkowski
Marc Paterno
Harrison Prosper
Antonio Sidoti

This group is charged with providing convenient interfaces to useful algorithms and software, in an experiment-independent manner, and to make these interfaces easily available to workshop participants. The group is responsible for documentation, bug-fixes and responding to participants' requests for enhancements. The group's guiding principles will be:

a) To re-use as much existing code as possible whether coded in C, C++ or FORTRAN and to develop new code for algorithms only if there is a compelling  reason to do so; for example, no code exists or the existing code is simply too opaque to warrant effort to understand it.

b) Interfaces should be as simple as possible and, to the degree possible,  uniform across analysis environments. The latter will be those actually used by physicists or that can be quickly learned (and remembered) by them with a minimum of effort.  The goal is to enable participants of the Run II Advanced Analysis Workhop (with varying levels of computer skills) to do real work, namely, apply advanced multivariate algorithms to specific physics topics in order to assess their potential for significantly improving physics analyses.

c) To build interfaces in simple (yet powerful) interpreted environments such as Python under a limited number of operating systems, for example, Linux and Windows.

d) Ideally, no compilation or linking should be necessary to use the interfaces. That is, the group will provide binaries in the form of shared object libraries (SOLs) under Linux or dynamic load libraries (DLLs) under Windows. (The group may, at some point, consider providing a web-based mechanism to allow workshop participants to create their own SOLs and DLLs from private C, C++ or FORTRAN codes. A user would upload source code and get back the corresponding SOL or DLL that could then be imported into, for example, Python.)
 

Here is an incomplete list of algorithms/software for which interfaces are needed:

Algorithms

Neural Networks (e.g, JETNET, MLPfit, SNN)

RaGS (Random Grid Search)

PDE (Probability Density Estimation)

PCA (Principal Component Analysis)

ICA (Independent Component Analysis)\

PC (Projection Pursuit)

Markov Chain Monte Carlo (for Bayesian analysis)

Bayesian Fitting


 

Utilities

Hbook

Root

Histoscope

Cernlib

Open Inventor (3D Graphics)

Simulation

Members

Alexander Belyaev
John Conway
Serban Protopopescu

This group is charged with providing simulated data in a convenient form, using tools provided by the Tools Group. We anticipate that the workshop participants will identify the set of reactions to be simulated and provide this group with specific requests. We anticipate providing ntuples that contain Monte Carlo simulations of Higgs, SUSY, Single Top and Leptoquark events, together with the most important backgrounds. The ntuples will contain both parton level as well as reconstruction level objects. Some mechanism, as yet unspecified, will be developed to allow the easy inclusion of event variables derived from the underlying event objects. The goal is to provide a uniform set of ntuples that all participants can somehow contribute to and share. However, in the short term the ntuples will probably contain event objects only.


PGS (Pretty Good Simulator). This code, which was developed during the Run II SUSY/Higgs Workshop, provides a fast simulation (thought to be accurate to about 15%) of  the response of typical collider experiments to high energy collision events. PGS includes an interface to PYTHIA and ISAJET. It will be extended to include HERWIG, and to allow the loading of 4-vectors generated with other programs such as CompHEP.

Evaluation

Members

Everyone!

(A list of specific projects and names will appear!)

Workshop participants will be encouraged to concentrate on a specific topic and try to apply to it a specific multivariate method. Each such effort should be documented and will eventually appear as part of a general workshop report.
 

Statistics

Members

Louis Lyons
Harrison Prosper
John Conway
Jim Linnemann
Pushpa Bhat

This group will focus on the statistical problems that arise from the workshop. However, from time to time it will also address some general issues that have bearing on all analyses. This group will also make available the results of deliberations of other similar groups withinhigh energy physics.


 
 


Last updated Feb. 15, 2001, Harrison B. Prosper and Pushpa Bhat