A New Version of the Price‘s Algorithm for Global Optimization

摘要: We present an algorithm for finding a global minimum of multimodal, multivariate function whose evaluation is very expensive, affected by noise and derivatives are not available. The proposed new version the well known Price‘s its distinguishing feature that it tries to employ as much possible information about objective obtained at previous iterates. has been tested on large set standard test problems shown satisfactory computational behaviour. used solve efficiently some difficult optimization deriving from study eclipsing binary star light curves.