| 摘要 |
PURPOSE:To efficiently find the optimum solution with fewer repeating times in a descent method. CONSTITUTION:A system to be an object is expressed by a variable vector X consisting of pieces of elements and quantity for evaluating the characteristic is expressed by an evaluation function vector F (X) consisting of the pieces of elements. At this time, solution is obtained so as to make the deviation of a target system evaluation function value min. and FT (X) F (X) is used as the evaluation measure. X where solution is a local min. value is obtained by solving (ATA+D) DELTAX=ATFO. D is a diagonal matrix correcting a nonlinear component so that an initial value d0i is given, repeating calculation is executed and the optimum value is obtained concerning the element di. As this d0i, the square value s12 and s22 of the singular value of (ATA) are obtained with respect to the Jacobi matrix A of F (X) and the repeating times can be reduced when the value is set between them. |
