Markov chain Monte Carlo scan of the pMSSM parameter space, in order to evaluate the SUSY-Yukawa Sum Rule. Based on work in arXiv:1201.5839.
UpsilonFit3 is written by Mike Saelim, and is a complete rewrite of the UpsilonFit2 code he wrote for arXiv:1201.5839. It is written in C++, and requires the GNU Scientific Library (GSL) and SuSpect, a FORTRAN program written to calculate SUSY spectra. It currently uses GSL version 1.16 and SuSpect version 2.43. Developed in NetBeans 7.4.
Interjection: The physics part of the code (package UpsilonFit3) is not done yet, but you can see plots of the MCMC code at work on toy problems, in ToyScan1_plots.pdf and ToyScan2_plots.pdf!
The MCMC algorithm is an adaptive Metropolis-Hastings algorithm with simulated annealing, based on the work in arXiv:hep-ph/0602187. It is adaptive because the step size changes to reflect the estimated size of the posterior distribution. Note that, while GSL has a simulated annealing package, we found it very limited.
It has three main parts:
-
The Mcmc package, which implements a general Markov chain Monte Carlo scan over a parameter space, given generic constraints. Class McmcScan is abstract, and requires a subclass to implement key functions.
-
The ToyScans package, which implements scans on small toy problems as tests/examples of the MCMC code at work.
-
The UpsilonFit3 package, which uses the Mcmc package to implement the scan over a subspace of the phenomenological Minimal Supersymmetric Standard Model (pMSSM) parameter space, constrained by a set of measurements made at the Large Hadron Collider (LHC) or a future linear electron-positron collider like the International Linear Collider (ILC). Its main class, PmssmScan, inherits from Mcmc::McmcScan. It also calculates the Upsilon parameter, defined in the SUSY-Yukawa Sum Rule (arXiv:1004.5350), for the resulting posterior distribution.
Mike Saelim would like to thank
- Maxim Perelstein for the original project and his guidance over the years, and
- Curran Muhlberger for his expert development and programming advice.
SuSpect is written by Abdelhak Djouadi, Jean-Loic Kneur, and Gilbert Moultaka: website, arXiv:hep-ph/0211331.