Abstract:In recent data mining research, the graph clustering methods, such as normalized cut and ratio cut, have been well studied and applied to solve many unsupervised learning applications. The original graph clustering methods are NP-hard problems. Traditional approaches used spectral relaxation to solve the graph clustering problems. The main disadvantage of these approaches is that the obtained spectral solutions could severely deviate from the true solution. To solve this problem, in this paper, we propose a new relaxation mechanism for graph clustering methods. Instead of minimizing the squared distances of clustering results, we use the l1-norm distance. More important, considering the normalized consistency, we also use the l1-norm for the normalized terms in the new graph clustering relaxations. Due to the sparse result from the l1-norm minimization, the solutions of our new relaxed graph clustering methods get discrete values with many zeros, which are close to the ideal solutions. Our new objectives are difficult to be optimized, because the minimization problem involves the ratio of nonsmooth terms. The existing sparse learning optimization algorithms cannot be applied to solve this problem. In this paper, we propose a new optimization algorithm to solve this difficult non-smooth ratio minimization problem. The extensive experiments have been performed on three two-way clustering and eight multi-way clustering benchmark data sets. All empirical results show that our new relaxation methods consistently enhance the normalized cut and ratio cut clustering results.

Discretize Relaxed Solution of Spectral Clustering via a Nonheuristic Algorithm

Discretize Relaxed Solution of Spectral Clustering via a Non-Heuristic Algorithm

A Novel and Effective Method to Directly Solve Spectral Clustering

Data Clustering Based on the Modified Relaxation Cheeger Cut Model

Discrete Nonnegative Spectral Clustering

A Unified Framework for Discrete Spectral Clustering.

Spectral Rotation versus K-Means in Spectral Clustering.

Large Graph Clustering with Simultaneous Spectral Embedding and Discretization

Discrete Multi-Graph Clustering.

Unified Spectral Clustering with Optimal Graph

New L1-Norm Relaxations and Optimizations for Graph Clustering.

Scalable Normalized Cut with Improved Spectral Rotation

Nonnegative Spectral Clustering with Discriminative Regularization.

Fast Optimization of Spectral Embedding and Improved Spectral Rotation

Spectral Clustering Via Half-Quadratic Optimization

Discrete and Balanced Spectral Clustering with Scalability

Spectral and Semidefinite Relaxation of the CLUHSIC Algorithm.

A Convex Formulation for Spectral Shrunk Clustering

Improved MinMax Cut Graph Clustering with Nonnegative Relaxation.

Spectral Clustering for Discrete Distributions

Discrete Optimal Graph Clustering