Clustering and Projected Clustering with Adaptive Neighbors

-We proposed a CAN clustering algorithm with adaptive neighbors, the learned similarity matrix can be directly used for clustering, without having to perform K-means or other discretization procedures.

Theoretical analysis reveals the proposed CAN clustering algorithm is connected with the K-means clustering problem, and the CAN can achieve much better clustering results than traditional K-means algorithm does.

For the high-dimensional clustering problem, we propose a Projected CAN (PCAN) algorithm, which performs clustering and dimensionality reduction simultaneously.

Theoretical analysis reveals the proposed PCAN clustering algorithm is connected with unsupervised LDA, and the PCAN can achieve better clustering or dimensionality reduction results than previous clustering algorithms or unsupervised dimensionality reduction algorithms do.