Dynamic Fuzzy c-Means (dFCM) Clustering and its Application to Calorimetric Data Reconstruction in High Energy Physics

TL;DR

This paper introduces the dynamic fuzzy c-means (dFCM) clustering algorithm, enhancing calorimetric data reconstruction in high energy physics by better handling non-uniformly distributed clusters compared to traditional FCM.

Contribution

The paper presents the dFCM algorithm, which dynamically generates and eliminates clusters, improving upon FCM for non-uniform data in calorimetric reconstruction.

Findings

dFCM outperforms FCM in non-uniform cluster scenarios

Both algorithms successfully applied to simulated calorimeter data

dFCM improves clustering accuracy and resolution

Abstract

In high energy physics experiments, calorimetric data reconstruction requires a suitable clustering technique in order to obtain accurate information about the shower characteristics such as position of the shower and energy deposition. Fuzzy clustering techniques have high potential in this regard, as they assign data points to more than one cluster,thereby acting as a tool to distinguish between overlapping clusters. Fuzzy c-means (FCM) is one such clustering technique that can be applied to calorimetric data reconstruction. However, it has a drawback: it cannot easily identify and distinguish clusters that are not uniformly spread. A version of the FCM algorithm called dynamic fuzzy c-means (dFCM) allows clusters to be generated and eliminated as required, with the ability to resolve non-uniformly distributed clusters. Both the FCM and dFCM algorithms have been studied and successfully applied to simulated data of a sampling tungsten-silicon calorimeter. It is seen that the FCM technique works reasonably well, and at the same time, the use of the dFCM technique improves the performance.