Probability Density Estimation (PDE) is a multivariate discriminationtechnique based on sampling signal and background densities defined by eventsamples from data or Monte-Carlo (MC) simulations in a multi-dimensional phasespace. In this paper, we present a modification of the PDE method that uses aself-adapting binning method to divide the multi-dimensional phase space in afinite number of hyper-rectangles (cells). The binning algorithm adjusts thesize and position of a predefined number of cells inside the multi-dimensionalphase space, minimising the variance of the signal and background densitiesinside the cells. The implementation of the binning algorithm PDE-Foam is basedon the MC event-generation package Foam. We present performance results forrepresentative examples (toy models) and discuss the dependence of the obtainedresults on the choice of parameters. The new PDE-Foam shows improvedclassification capability for small training samples and reduced classificationtime compared to the original PDE method based on range searching.
展开▼
