TEMPORAL LOW RANK SUBSPACE CLUSTERING BASED ON ITERATED WEIGHTED SINGULAR VALUE

In order to improve the application scope and clustering performance, a temporal low rank subspace clustering algorithm based on iterative weighted singular value is proposed. In order to solve the problem that the constraint based on kernel norm usually leads to sub-optimal solution, a strong convex optimization method was proposed for the cost function, which guaranteed the unique solution of the subsequent update sub-problem theoretically. An iterative weighted singular value minimization algorithm and a singular value threshold acceleration algorithm were proposed by introducing extrapolation technique and rank progressive operation, so as to reduce the computational complexity and ensure fast convergence. Experimental results on several open data sets show that the model can reveal the internal structure of data, extend the application scope and improve the clustering performance.