METHOD FOR CLUSTERING DATA BASED CONVEX OPTIMIZATION

摘要

A data clustering method based on a convex optimization method is provided to obtain a more ideally optimal feasible solution by using a matrix to which a strong duality for a graph partition method of a user is well reflected in a semi-definite relaxation process to improve graph spectral clustering performance. An optimal feasible solution which satisfies a provided strong duality is obtained by using a convex optimization method for an inputted objective function(S1-S3). Eigen-values are extracted and clustered from the optimal feasible solution(S4). A semi-definite relaxation method is used as the convex optimization method.