sbein / rgs

Home of the Random Grid Search algorithm. A very simple, but surprisingly effective, way to find rectangular cuts. Developed by yours truly, Chip Stewart, Pushpa Bhat and generalized by Sezen Sekmen.

RGS

Introduction

The Random Grid Search (RGS) algorithm is a simple, but surprisingly effective, way to find rectangular cuts. Developed by Harrison Prosper, Chip Stewart, Pushpa Bhat and generalized by Sezen Sekmen to include two-sided and staircase cuts. A two-sided cut is a cut of the for (x1 < x < x2), while a staircase cut is the OR of two or more one-sided cuts.

Installation

This package depends on the package Root from CERN. To install Root follow the instructions at the Root website. Then do

git clone https://github.com/hbprosper/RGS.git
cd RGS
make
source setup.sh

The setup need be done only once per terminal session. (Note: the bash setup can be "sourced" from any directory, but currently the non-bash setup must be sourced from the RGS directory.)

Examples

There are two examples in the examples directory of RGS. These examples require Root data files, which can get obtained using the commands

cd examples/data
wget http://www.hep.fsu.edu/~harry/RGS/data/Higgs.tar.gz 
tar zxvf Higgs.tar.gz

or downloaded from the website via a web browser. Use a similar procedure for the SUSY data files Susy.tar.gz.

Higgs

This example illustrates three RGS optimizations, HO1, HO2, and HO3, designed to enhance the ratio of Higgs vector boson fusion (VBF) events to Higgs gluon gluon fusion (ggF) events and di-Z boson events. Each optimization can be run by executing the train.py program followed by analysis.py program. For example, HO1 can be run as follows

cd examples/Higgs/HO1
./train.py

which will run RGS and store its results in a file called HO1.root. To analyze the results of RGS do

./analysis.py

which will read the results from HO1.root and write the results to r_HO1.txt and also produce a couple of plots.

SUSY

This example illustrates three optimizations, SO1, SO2, and SO3 that use staircase cuts to improve the search for SUSY events. Switch to the SUSY directory and proceed as in the Higgs example.