Study On Manifold Spectral Clustering

Clustering research is an important part of high dimensional data analysis.Most of the traditional clustering methods are based on distance criterion.As big data comes,data patterns vary widely and data structure is intricate.In this sisuation we can not find the appropriate distance criteria.The introduction of data geometry has become a hot spot for researchers,and spectral clustering is one of the popular clustering methods.The spectral clustering utilizes the geometric structure of the data to realize the transformation of the high-dimensional space to the low-dimensional subspace.It is widely used in the actual scene of face recognition and image segmentation.According to the composition of the spatial structure of the data is generally divided into linear subspace clustering and nonlinear subspace clustering.The manifold learning is used to solve the problem of subspace superposition and crossover in nonlinear spatial clustering.Multifractal spectral clustering algorithm is one of the most bulge clustering algorithms.This paper mainly introduces the basic concepts and methods in clustering research,including the principal component analysis(PCA)reduction method and the basic k-means clustering method commonly used in high dimensional space.In this paper,we introduce the spectral clustering algorithm of manifold in recent years.Among the linear subspace clustering problem,this paper introduces the sparse subspace clustering algorithm(SSC)in spectral clustering,and through e.xperiments,The effectiveness of the method can solve the subspace clustering problem under linear mixed space,such as intersecting linear clustering problem,which solves the problem that the traditional K-means can not solve.On the other hand,the study also finds that when the data is sampled in a non-linear subspace in a high dimensional space,the clustering method based on linear subspace is not only computationally computationally efficient but also has no noise and parameters More sensitive,the model is not good enough robustness.SMMC al-gorithm is also a kind of spectral clustering algorithm,mainly to solve the nonlinear subspace clustering problem.Based on the SMMC algorithm and some practical prob?lems,this paper puts forward some improvements,which solves the problem of motion segmentation and face recognition.Based on the actual data,the above algorithm is compared experimentally,and the adaptive algorithm is analyzed.It is proved that the spectral clustering algorithm based on multi-manifold learning has a good effect in the cross-nonlinear subspace clustering.Finally,in practice,based on the multi-manifold clustering method,Variable clustering,motion scene proposed based on the displace-ment of the clustering method.It can be seen from the experimental results that the clustering accuracy of the model is higher than that of the previous method.