Similarity Measure Based on Adaptive Neighbors for Spectral Clustering

Abstract

Spectral clustering has become one of the most popular clustering methods for exploratory data analysis. Similarity measure is crucial to the performance of spectral clustering. In this paper, to improve spectral clustering, we propose an efficient method that measures the similarity for spectral clustering by considering the adaptive and optimal neighbors of each data based on the local structure. The similarity of two data points is measured based on the probability that these two data points are neighbors. The proposed method is able to explore the underlying similarity relationships between data points, and is robust to the datasets with high dimensions. We evaluate the proposed method by using high dimensional real-world datasets. The experimental results demonstrate that the proposed method not only achieves good performance, but also outperforms the traditional spectral clustering algorithms.