TerraFerMA A Suite of Multivariate Analysis Tools Sherry Towers SUNY-SB smjt@fnal.gov Today I will be talking to you about work I have done over the past couple of years in the field of multivariate analysis techniques. We will begin by briefly summarising the strengths and weaknesses of various common multivariate techniques. We will then discuss the development of GEM, a multivariate tool I developed as a graduate student…as we will see, GEM is quite similar to the PDE multivariate approach, with a few differences, which will become apparent as we discuss the performance of GEM. Now, a little talked about problem in high energy physics today is the proliferation of the number of variables used in analyses…the last decade or so has witnessed a sharp rise in the popularity of neural networks in physics analyses. Neural nets are a very powerful and valuable tool. Unfortunately, the ease of adding discriminators to a neural net has led to some analyses with literally dozens of variables. This is a phenomenon I call “overloading the dimensionality of an analysis”. I’ll talk about how this can be detrimental, and we will discuss a simple step-by-step method that can be used to avoid this problem. I will then show the result of applying the variable-reduction method to tauID and topID At D0. I’ll also show in these two examples how the GEM algorithm performs compared to the neural networks previously used in these analysis. Version 1.0 has been released useable by anyone with access to the CLHEP and Root libraries www-d0.fnal.gov/~smjt/multiv.html

TerraFerMA=Fermilab Multivariate Analysis (aka “FerMA”) TerraFerMA is, foremost, a convenient interface to various disparate multivariate analysis packages (ex: MLPfit, Jetnet, PDE/GEM, Fisher discriminant, binned likelihood, etc) User first fills signal and background (and data) “Samples”, which are then used as input to TerraFerMA methods. A Sample consists of variables filled for many different events.

Using a multivariate package chosen by user (ie; NN’s, PDE’s, Fisher Discriminants, etc), TerraFerMA methods yield probability that a data event is signal or background. TerraFerMA also includes useful statistics tools (means, RMS’s, and correlations between the variables in a Sample), and a method to detect outliers.

TerraFerMA makes it trivial to compare performance of different multivariate techniques (ie; simple to switch between using a NN and a PDE (for instance) because in TerraFerMA both use the same interface!) TerraFerMA makes it easy to reduce the number of discriminators used in an analysis (optional TerraFerMA methods sort variables to determine which have best signal/background discrimination power) TerraFerMA web page includes full documentation/descriptions

Future Plans… To package TerraFerMA with ROOT, dependencies on CLHEP matrix methods and random number generators must be excised. Process has begun. Predicted completion date: December. In the meantime, FerMA is fully useable/interfaceable with root-tuples by using it in compiled mode. See the package and users’ guide for detailed examples.

TerraFerMA documentation: TerraFerMA Version 1.0 TerraFerMA documentation: www-d0.fnal.gov/~smjt/ferma.ps TerraFerMA users’ guide: www-d0.fnal.gov/~smjt/guide.ps TerraFerMA package: …/ferma.tar.gz (includes example programs)