Simultaneous Dempster-Shafer clustering and gradual determination of number of clusters using a neural network structure

TL;DR

This paper introduces a neural network-based method that simultaneously clusters evidence and determines the optimal number of clusters within the Dempster-Shafer framework, improving flexibility and accuracy.

Contribution

It extends previous fixed-cluster methods by enabling dynamic determination of cluster count during neural network iteration.

Findings

Effective simultaneous clustering and cluster number determination.

Gradual feedback improves clustering accuracy.

Method outperforms fixed-cluster approaches.

Abstract

In this paper we extend an earlier result within Dempster-Shafer theory ["Fast Dempster-Shafer Clustering Using a Neural Network Structure," in Proc. Seventh Int. Conf. Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU'98)] where several pieces of evidence were clustered into a fixed number of clusters using a neural structure. This was done by minimizing a metaconflict function. We now develop a method for simultaneous clustering and determination of number of clusters during iteration in the neural structure. We let the output signals of neurons represent the degree to which a pieces of evidence belong to a corresponding cluster. From these we derive a probability distribution regarding the number of clusters, which gradually during the iteration is transformed into a determination of number of clusters. This gradual determination is fed back into the neural structure at each iteration to influence the clustering process.