Abstract:

The key technique of improving sequential approximation modeling accuracy and convergence efficiency is to establish reasonable and efficient sequential sampling criteria. To obtain target system information effectively and accelerate convergence of modeling accuracy, a multicriterion sequential modeling method based on partial cross validation is proposed. By combination of partial cross validation error estimation criterion, maximum sequential modeling accumulation change criterion and maximum entropy criterion, the spacefilled sampling and irregular region importance sampling are realized, and approximation accuracy and convergence efficiency are improved. Test examples validate the feasibility and efficacy of this method.