Nonlinear discriminative embedding for clustering via spectral regularization

Authors:
Yubin Zhan;Jianping Yin
Affiliations:
School of Computer, National University of Defense Technology, Changsha, P.R. China;School of Computer, National University of Defense Technology, Changsha, P.R. China
Venue:
PAKDD'11 Proceedings of the 15th Pacific-Asia conference on Advances in knowledge discovery and data mining - Volume Part I
Year:
2011

Citing 9
Cited 0

Normalized Cuts and Image Segmentation

IEEE Transactions on Pattern Analysis and Machine Intelligence
Modeling the Shape of the Scene: A Holistic Representation of the Spatial Envelope

International Journal of Computer Vision
Laplacian Eigenmaps for dimensionality reduction and data representation

Neural Computation
Multiclass Spectral Clustering

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Semi-Supervised Learning on Riemannian Manifolds

Machine Learning
Graph Embedding and Extensions: A General Framework for Dimensionality Reduction

IEEE Transactions on Pattern Analysis and Machine Intelligence
A tutorial on spectral clustering

Statistics and Computing
Spectral embedded clustering

IJCAI'09 Proceedings of the 21st international jont conference on Artifical intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we propose a novel nonlinear discriminative dimensionality reduction method for clustering high dimensional data. The proposed method first represents the desired low dimensional nonlinear embedding as linear combinations of predefined smooth vectors with respect to data manifold, which are characterized by a weighted graph. Then the optimal combination coefficients and optimal cluster assignment matrix are computed by maximizing between-cluster scatter and minimizing within-cluster scatter simultaneously as well as preserving smoothness of the cluster assignment matrix with respect to the data manifold. We solve the optimization problem in an iterative algorithm, which is proved to be convergent. The contributions of the proposed method are two folds: 1) obtained nonlinear embedding can recover intrinsic manifold structure as well as enhance the cluster structure of the original data; 2) the cluster results can also be obtained in dimensionality reduction procedure. Extensive experiments conducted on UCI data sets and real world data sets have shown the effectiveness of the proposed method for both clustering and visualization high dimensional data set.