化学计量学非线性偏最小二乘算法进展评述

评述了偏最小二乘法(PLS)近年来非线性化的进展以及重要算法的原理、特性和应用。一类是对自变量进行非线性前处理的方法，包括二次变换、Chebychev变换和基于机理的变换；另一类是基于内部非线性映射的PLS，主要有多项式映射、样条函数映射、由遗传算法辅助实施的任意非线性映射和神经网络映射；介绍了具有类似功能的正交网模型。本文还就算法的普适性、稳健性和简易性，及其非线性化的演变进程予以评述、总结和展望。

Recent Development of Non-linear Patital Least Squares in Chemometrics

In this paper we review the recent progresses of non linear partial least squares (PLS) applied in the field of chemometrics. The methods using transformed original variables include squared X PLS, chebychev PLS and Squared X PLS based on mechanism. Inner non linear algorithm such as polynomial PLS, spline PLS, Numeric Genetic Algorithm PLS, and neural network PLS are slso mentioned. Orthogonal latent variable neural network (orthonet), another method similar to non linear PLS, are also mentioned. The principles, characters, and applications of these methods are introduced and discussed. The generalization, stability, simplicity and evolving progresses of these algorithms are evaluated, summarized and prospected.