Dynamic sample point selecting method based on relativity principle

Abstract

The invention discloses a dynamic sample point selecting method based on the relativity principle. According to the method, based on the relativity principle, according to the initially-selected sample point information, leave-one-out errors of all sample points are calculated, for the sample points with the large leave-one-out errors, a variation function is used for calculating the influence range of the leave-one-out errors in each variable direction, then the plus point area range of the sample point in each variable direction is determined, at last, sample points are additionally selected in each plus point area, and therefore targeted selecting of the sampling points is achieved. According to the dynamic sample point selecting method based on the relativity principle, the selecting interval of the sample points is guided by fully using the characteristics of performance space, so that the 'distribution according to needs' of the sample points is achieved. The demand for the sample point distribution of function space is considered on the sample point selecting aspect, the precision of a corresponding approximation model is obviously higher than the precision of an approximation model corresponding to one-time sample point selecting, and the utilization efficiency of the sample points is higher.

Description

technical field [0001] The invention relates to a sample selection method, in particular to a dynamic sample point selection method based on the principle of correlation. Background technique [0002] Since the numerical calculation of ship type performance is time-consuming, and hundreds of ship type schemes need to be calculated in ship type optimization, in order to improve the efficiency of ship type optimization and realize its engineering application, it is often necessary to construct an approximate model of ship type related performance. In the process of constructing the approximate model, the selection of sample points is very important, because it provides the data basis for the construction of the approximate model. Each sample point here represents a scheme, that is, a ship type scheme. The traditional approach is based on the variation range of the variable, in the variable space G s Evenly select the sample point P in n (P 1 , P 2 ,…P n ), and then calcu...

AI Technical Summary

Problems solved by technology

At this time, using uniform design to select sample points will have the problem of unreasonable "distribution" of sample points, and the "as-needed distribution" of sample points has not been realized.

Images