Efficient Monte Carlo clustering in subspaces

摘要

Clustering of high-dimensional data is an important problem in many application areas, including image classification, genetic analysis, and collaborative filtering. However, it is common for clusters to form in different subsets of the dimensions. We present a randomized algorithm for subspace and projected clustering that is both simple and efficient. The complexity of the algorithm is linear in the number of data points and low-order polynomial in the number of dimensions. We present the results of a thorough evaluation of the algorithm using the OpenSubspace framework. Our algorithm outperforms competing subspace and projected clustering algorithms on both synthetic and real-world data sets.