manet

Manet is an implementation for NVIDIA GPUs of a statistical two-sample comparison test known as the energy test.

This implementation has been created to facillitate the use of the test in searches for CP violation in the phase-space of multi-body decays but the implementation is general and can be used in other applications. The energy test requires the computation of the distances between all elements of the samples. This implementation for GPUs may be useful in the case that large sample sizes (greater than 10⁵ elements) are being compared.

Energy Test

The energy test is a multivariate two sample test described in:

B. Aslan and G. Zech, New test for the multivariate two-sample problem based on the concept of minimum energy, J. Stat. Comput. Simul. 75 (2005) 109.

B. Aslan and G. Zech, Statistical energy as a tool for binning-free, multivariate goodness-of -t tests, two-sample comparison and unfolding, Nucl. Instrum. Meth. A537 (2005) 626

The application of this test to CP violation in multibody decays and this GPU implementation is described in:

Chris Parkes, Shanzhen Chen, Jolanta Brodzicka, Marco Gersabeck, Giulio Dujany, William Barter (Manchester U.), On model-independent searches for direct CP violation in multi-body decays, Dec 14, 2016. 16 pp. e-Print: arXiv:1612.04705

NVIDIA GPUs and CUDA

Manet runs on NVIDIA GPUs, it requires the CUDA platform and utilises the THRUST library, please see:

NVIDIA Corporation, Thrust quick start guide, DU-06716-001 (2014).

If your graphics card is from NVIDIA and it is listed in NVIDIA graphics cards, your GPU is CUDA-capable.

The NVIDIA CUDA Toolkit is available at Toolkit dowload, and can be installed following the instructions in Installation instructions.

Installing and running the code

The Manet program can be downloaded using the link on the left. If you wish to run locally, before running the program, first make sure you have CUDA-capable GPU in your computer and CUDA Tookit installed. To verify that the GPU is CUDA-capable, i.e., NVIDIA GPU, you could do

$ lspci | grep -i nvidia

If you see the list of NVIDIA GPUs, your computer is CUDA-capable. Before running Manet, set the environment variables by

$ export PATH=/usr/local/cuda/bin:$PATH

$ export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

then install Manet into your computer by executing

$ sh Manet-x.x.x-Linux.sh

To get help of the program, you can do

$ ./exeMANET -h

About

This code has been developed by a team in the particle physics group at the University of Manchester (MANchester Energy Test - Manet) and is made freely available for non-commercial applications. Please cite

Chris Parkes, Shanzhen Chen, Jolanta Brodzicka, Marco Gersabeck, Giulio Dujany, William Barter (Manchester U.), On model-independent searches for direct CP violation in multi-body decays, Dec 14, 2016. 16 pp. e-Print: arXiv:1612.04705

if you utilise this code or concepts.

For questions please contact Shanzhen.Chen ATNOSPAM cern.ch.

This page was last edited in April 2017.