| 摘要 |
A similarity measurement method comprises: acquiring directional relations among nodes in a network and determining a transfer matrix (101); computing a constraint matrix according to the transfer matrix and an attenuation factor acquired (102); constructing a system of linear equations, wherein the coefficient matrix of the system of linear equations is the constraint matrix, the variable of the system of linear equations is a correction vector (103); solving the system of linear equations by iteration according to a Jacobi method, and determining the correction vector (104); and computing similarities among the nodes according to the transfer matrix, the attenuation factor and a diagonal correction matrix generated from the correction vector (106). In the method, the Jacobi method is utilized to determine the correction vector and thus similarities among the nodes can be further computed. Various elements of the correction vector are computed independently of one another in every iteration of the Jacobi method and in this way parallel computation can be achieved. Thus, the time for computation can be effectually shortened by use of computer cluster and the time complexity and space complexity during computation can be reduced; in addition, the method can apply to large networks. |
