DEVICE AND METHOD FOR GENERATING LINEAR CONSTRAINT CONDITION, DEVICE FOR SOLVING POSITIVE SEMIDEFINITE OPTIMIZATION PROBLEM, METRIC LEARNING DEVICE, AND COMPUTER PROGRAM

摘要

Disclosed is a linear constraint condition generating device for accurately and quickly deriving a solution, during approximation of a linear constraint problem to solve a positive semidefinite programming problem. This linear constraint condition generating device is equipped with: a matrix operation means for deriving a matrix in which are arrayed coefficients representing results of linear approximation of a target variable of a semidefinite programming problem, and a transposed matrix of this matrix, as well as calculating the sum of the matrix and the transposed matrix; and a vector operation means for calculating characteristic values and characteristic vectors in relation to a matrix representing operation results of the matrix operation means, as well as selecting a specific characteristic vector from among the operation results, and employing the information represented by the specific characteristic vector as information representing the linear constraint condition of the positive semidefinite programming problem.