Fuzzy clustering is one of the most popular approaches and has attracted considerable attention in many fields. However, high computational cost become a bottleneck which limits its applications large-scale problems. Moreover, fuzzy algorithms are sensitive to noise. To address these issues, novel algorithm, called fast based on anchor graph (FFCAG), proposed. The FFCAG algorithm integrates anchor-based similarity construction membership matrix learning into unified framework, such that...

Fuzzy clustering algorithms have been widely used to reveal the possible hidden structure of data. However, with increasing data amount, large scale has brought genuine challenges for fuzzy clustering. Most suffer from long time-consumption problem since a amount distance calculations are involved update solution per iteration. To address this problem, we introduce popular anchor graph technique into and propose scalable algorithm referred as Scalable Clustering Anchor Graph (SFCAG). The...

Spectral clustering have attracted more and attention due to their well-defined mathematical frameworks superior performance. However, there still exist two limitations be solved: 1) most spectral methods consist of independent stages, which may cause unpredictable deviation obtained results from the genuine ones lead severe information loss performance degradation; 2) employ hard mode, lacks interpretability for data points in boundary area belonging multiple clusters. To simultaneously...

Bipartite spectral graph partitioning (BSGP) method as a co-clustering method, has been widely used in document clustering, which simultaneously clusters documents and words by making full use of the duality between words. It consists two steps: 1) construction 2) singular value decomposition on bipartite to compute continuous cluster assignment matrix, followed post-processing get discrete solution. However, generated is unstructured fixed. heavily relies quality construction. Moreover,...

Spectral clustering has been attracting increasing attention due to its well-defined framework and excellent performance. However, most traditional spectral methods consist of two separate steps: 1) Solving a relaxed optimization problem learn the continuous labels, 2) Rounding labels into discrete ones. The results relax-and-discretize strategy inevitably result in information loss unsatisfactory Moreover, similarity matrix constructed from original data may not be optimal for since usually...

Purpose The purpose of this paper is to focus on the design automated fabric defect detection based cascaded low-rank decomposition and maintain high quality control in textile manufacturing. Design/methodology/approach This proposed a algorithm decomposition. First, constructed Gabor feature matrix divided into sparse using technique, used as priori where higher values indicate probability abnormality. Second, we conducted second for texton under guidance matrix. Finally, an improved...

Fabric defect detection plays a curial step in the quality control of textiles. Existing fabric methods are lack adaptability and have poor performance. A novel method based on Gabor filter tensor low-rank recovery was proposed this paper. Defect-free images specified direction, while defects damage their regularity direction. Therefore, direction feature is for detection. For different kinds image, information also distinct. In order to characterize all we adopted bank directional filters...

The use of random and unique characteristics the fiber distribution is an effective anti-fake method in modern industry.Detecting combing with from images automatically can improve efficiency industrial production greatly.In this paper, we propose a frequency salient based detection which robust to light condition slightly background impurity.Our exploits low level features color luminance, easy implement, fast, provides full resolution.The resulting saliency maps are better suited...