NN-EVCLUS: Neural network-based evidential clustering

Evidential clustering is an approach to based on the use of Dempster-Shafer mass functions represent cluster-membership uncertainty. In this paper, we introduce a neural-network evidential algorithm, called NN-EVCLUS, which learns mapping from attribute vectors functions, in such way that more similar inputs are mapped output with lower degree conflict. The neural network can be paired one-class support vector machine make it robust outliers and allow for novelty detection. trained minimize discrepancy between dissimilarities degrees conflict all or some object pairs. Additional terms added loss function account pairwise constraints labeled data, also used adapt metric. Comparative experiments show superiority N-EVCLUS over state-of-the-art algorithms range unsupervised constrained tasks involving both dissimilarity data.