Method and apparatus for selecting integrated circuit device neural network modeling sample

Abstract

A method and apparatus for selecting an integrated circuit device neural network modeling sample; for an input variable having the largest mean impact value, by means of continually and equally dividing an interval of the input variable, until relative errors of all divided intervals are equal to or less than a preset error precision, only at which point the equal division action stops, and the length of the divided interval having the smallest length being taken as a step length of the output variable; the step lengths of other input variables then being respectively calculated, according to the step length of the input variable; and finally, for each input variable, points being extracted according to a change interval and the step length thereof, thereby obtaining a sample point set of each input variable. Thus, a low amount of sample data may be selected under the condition of a given precision, the low sample data amount further saving testing expenditures required by device modeling, and increasing the training speed of a neural network.