Multi-view Clustering using Barycentric Coordinate Representation

We consider clustering issues where the available attributes can be divided into various independent groups that frequently offer complimentary information. We concentrate on real-world applications in this paper where a single instance can be represented by a number of heterogeneous features. As wa...

Full description

Saved in:
Bibliographic Details
Main Authors: QIAN, Xiaotong, JIN, Lili, CABANES, Guenael, RASTIN, Parisa, GROZAVU, Nistor
Format: Conference Proceeding
Language:English
Subjects:
barycentric coordinate
Clustering algorithms
Complexity theory
Data models
Measurement
multi-view clustering
Neural networks
Prototypes
Time complexity
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We consider clustering issues where the available attributes can be divided into various independent groups that frequently offer complimentary information. We concentrate on real-world applications in this paper where a single instance can be represented by a number of heterogeneous features. As was performed successfully in the prior work of clustering on a single view dataset by using barycentric coordinate(BC) representation, and also a recent KMeans-based multi-view clustering RMKMC which proposed that the weights of views can be auto-updated by introducing a hyperparameter γ, we further propose a novel approach of multi-view clustering BCmvlearn by combining these two approaches to reduce complexity without sacrificing clustering quality. In addition, the vector form of the original dataset not being absolutely necessary due to the distance-based property of the BC representation, a variant application of multi-modal clustering is also achievable.
ISSN:2161-4407
DOI:10.1109/IJCNN54540.2023.10191546