Nonnegativity-enforced Gaussian process regression  

作　　者：Andrew Pensoneault Xiu Yang Xueyu Zhu 

机构地区：[1]Department of Mathematics,University of Iowa,Iowa,IA 52246,USA [2]Department of Industrial and Systems Engineering,Lehigh University,Bethlehem,PA 18015,USA

出　　处：《Theoretical & Applied Mechanics Letters》2020年第3期182-187,共6页力学快报（英文版）

基　　金：supported by Simons Foundation;supported by the U.S. Department of Energy Office of Science, Office of Advanced Scientific Computing Research as part of Physics-Informed Learning Machines for Multiscale and Multiphysics Problems (PhILMs)

摘　　要：Gaussian process(GP)regression is a flexible non-parametric approach to approximate complex models.In many cases,these models correspond to processes with bounded physical properties.Standard GP regression typically results in a proxy model which is unbounded for all temporal or spacial points,and thus leaves the possibility of taking on infeasible values.We propose an approach to enforce the physical constraints in a probabilistic way under the GP regression framework.In addition,this new approach reduces the variance in the resulting GP model.

关 键 词：Gaussian process regression Constrained optimization 

分 类 号：O212.1[理学—概率论与数理统计]