A Gaussian process-based Incremental Neural Network for Online Clustering
Xiaoyu Wang, Jun‐ichi Imura · 2019
This paper proposes a two-stage online clustering algorithm. First, it generates cluster prototypes using a Gaussian process-based Incremental Neural Network (GPINN), where 1) the network structure is updated in an online adaptive mode and 2) both combinatorial effects and the similarity between the weight vectors of nodes are considered. Second, clusters are detected by constructing the minimum spanning tree of GPINN. Besides, some of its properties are discussed. The experimental results on both synthetic data and real-world data show that our method achieves remarkable improvement in clustering accuracy compared with previous incremental neural network algorithms.