自适应链接超平面模型(AHH)是一种自适应的分片线性模型,可以作为一种人工神经网络用于非线性函数逼近。通过代数等价变换,该文证明,基于单纯形划分的高阶典范模型(HL-CPWL)的基函数等价于AHH模型的一种基函数,HL-CPWL模型是AHH模型的一个特例。较之HL-CPWL模型,AHH模型的定义域划分更为灵活,使得其更适合于函数逼近。AHH的通用逼近性也由HL-CPWL具有通用逼近能力而直接得到。仿真结果表明,较之HL-CPWL,AHH能够以较少的参数给出较好的逼近结果,具有更好的模型质量。 Adaptive Link Hyperplane Model (AHH) is an adaptive piecewise linear model that can be used as an artificial neural network for nonlinear function approximation. By algebraic equivalence transformation, the paper shows that the basis function of HL-CPWL is equivalent to a kind of basis function of AHH model. HL-CPWL model is a special case of AHH model. Compared with the HL-CPWL model, the domain definition of AHH model is more flexible, making it more suitable for function approximation. The general approximation of AHH is also obtained directly by HL-CPWL with universal approximation ability. Simulation results show that, compared with HL-CPWL, AHH can give better approximation results with fewer parameters and have better model quality.