Abstract:Spectral clustering is a very popular graph-based clustering technique that partitions data groups based on the input data similarity matrix. Many past studies based on spectral clustering, however, do not consider the global discriminative structure of the dataset. Also, the benefits of using more than one kernel have not been fully exploited with respect to spectral clustering, although it has been established by these past studies that using more than one kernel in clustering can result in a more accurate clustering than those obtained with a single kernel. Multi-kernel approaches, however, tend to be more time consuming compared to single kernel methods. To compensate these drawbacks, we integrate a global discriminative term into the clustering with an adaptive neighbor framework. This is done to preserve both the global geometric information and global discriminative information in a dual kernel space, in an attempt to optimize clustering performance. Via co-regularization, we utilize more than one kernel space to take advantage of the benefits of multiple kernels. We, however, use two heterogeneous kernels to help us reduce clustering time, since the ability to quickly process data is as equally important as its accuracy in this era of information explosion. Since these different kernel spaces admit the same underlying clustering of the data, we approach the problem looking for clustering consistent across the two kernel views. Hence we are able to detect the non-linear intrinsic geometrical information of the dataset. We perform clustering using the obtained indicator matrix from our modified Laplacian utilizing k-means. Our Experimental outcomes show that our approach gives satisfactory results in terms of accuracy and NMI, with time-to-cluster savings in comparison to other state-of-the-art clustering methods using both synthetic and public datasets.

Nonlinear discriminant clustering based on spectral regularization

Nonnegative Spectral Clustering with Discriminative Regularization.

Global discriminative-based nonnegative spectral clustering.

Robust Discriminative Clustering with Sparse Regularizers

Clustering-Based Nonlinear Dimensionality Reduction On Manifold

Discriminative Nonnegative Spectral Clustering with Out-of-Sample Extension

Reweighted Non-convex Non-smooth Rank Minimization Based Spectral Clustering on Grassmann Manifold

Dimension Reduction of Hyperspectral Images with Sparse Linear Discriminant Analysis

Sparse and Low-Rank Subspace Data Clustering with Manifold Regularization Learned by Local Linear Embedding

Co-Regularized Discriminative Spectral Clustering with Adaptive Similarity Measure in Dual-Kernel Space

Linear discriminant analysis with spectral regularization

A Dimension Reduction Algorithm Preserving Both Global and Local Clustering Structure

A Convex Formulation for Spectral Shrunk Clustering

A Spectral Clustering on Grassmann Manifold Via Double Low Rank Constraint

Adaptive Local Learning Regularized Nonnegative Matrix Factorization for Data Clustering.

Learning Regularized LDA by Clustering.

Discriminative K-Means Laplacian Clustering

Nonparametric Discriminant Multi-Manifold Learning for Dimensionality Reduction

Discrete Nonnegative Spectral Clustering

Robust Discriminant Subspace Clustering with Adaptive Local Structure Embedding.

Dual hybrid manifold regularized non-negative matrix factorization with discriminability for image clustering