Answers are grounded in this paper and cite their sources.
In high-dimensional data, clusters lay on subspaces of the original space. Projected clustering is challenging because the circular dependence on the subspaces searching and the cluster assignment. Without extra information, it seems impossible to break the circular dependence. In this paper, we propose Semi-supervised Projected Subspace Clustering to utilize pairwise constraints to break the circular dependence. The algorithm consists of two phases: subspace finding and cluster forming. In the first phase, constraints are mobilized to find subspaces based on the feature correlation. In the second phase, an adaptive dimension voting is employed to refine the subspaces and a Semi-supervised k-medoid-like assign method is adopted to detect clusters. We emphasize the distance divergence in the process of adaptive dimension voting guided by constraints. Experiments on real world datasets show that the three variants of SSPSC outperform baseline algorithms in accuracies and scalability.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2023) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: