Gravitational Clustering: A simple, robust and adaptive approach for distributed networks
Abstract
New gravitational clustering method to adaptively estimate the number of clusters based on a set of features. Key idea: exploit gravitational force between mass units to determine regions of highest density of feature vectors, i.e. the clusters. Required for object/source labelling under the multi device multi task paradigm. Cooperative and distributed cluster enumeration achieved by diffusion-adaptation scheme. Application to distributed cooperative multi-view camera networks shows the applicability to real-world problems.
- Publication:
-
Signal Processing
- Pub Date:
- August 2018
- DOI:
- 10.1016/j.sigpro.2018.02.034
- arXiv:
- arXiv:1709.02287
- Bibcode:
- 2018SigPr.149...36B
- Keywords:
-
- Adaptive distributed clustering;
- Cluster enumeration;
- Robust;
- Outlier;
- Multi device multi task (MDMT);
- Wireless sensor networks;
- Labelling;
- Computer Science - Distributed;
- Parallel;
- and Cluster Computing;
- Computer Science - Computer Vision and Pattern Recognition
- E-Print:
- 12 pages, 9 figures