Abstract:Adaptive hinging hyperplanes (AHH) is a popular continuous piecewise linear (CPWL) model. It has been proved that any continuous nonlinear function can be approximated by a CPWL function with arbitrary precision. The existing identification of AHH simply traverses all the dimensions on the pre-given splitting points to select the best, which fails to consider all the parameters synchronously and the randomness in the splitting, thus the identified model may not be optimal. In this paper, we propose an improved method to identify AHH model with coordinate update strategy. We first use the existing identification method of AHH to initially obtain a basic model structure, and afterwards alternatively optimize the parameters to improve accuracy. Specifically, to explore the interactive and global effects among all the nonlinear parameters, adaptive block coordinate DIRECT (ABCD) algorithm is employed to simultaneously optimize the nonlinear parameters, while the linear parameters can be calculated by least squares (LS) method. Besides, the proposed method is promising to conduct extensions to identify different CWPL models or other nonlinear models even with various error criteria. Numerical experiments show that the proposed method improves the accuracy and stability in identifying AHH and it can even achieve higher accuracy with simpler model structure.

Identification Algorithm for Standard Continuous Piecewise Linear Neural Network

Nonlinear system identification with continuous piecewise linear neural network

Continuous piecewise linear identification with moderate number of subregions

Approximation And Inverse Control Of Nonlinear System Using Standard Continuous Piecewise Linear Neural Networks

Configuration of Continuous Piecewise-Linear Neural Networks

A special kind of neural networks: continuous piecewise linear functions

Closed-Loop Subspace Identification Based on Kpls

On the Number of Linear Regions of Convolutional Neural Networks with Piecewise Linear Activations.

A New Neural Network Model And Its Application

Constructive PWL Neural Network Approximation

Sequential Global Linear Programming Algorithm for Continuous Piecewise Linear Programming

Standard neural network model and its application

Continuous Piecewise Linear Programming Via Concave Optimization and Genetic Algorithm

An Improved Coordinate Update Method For The Identification Of Adaptive Hinging Hyperplanes Model

Learning continuous piecewise non-linear activation functions for deep neural networks

Nonlinear System Identification with Piecewise-Linear Functions

Training neural networks for solving 1-D optimal piecewise linear approximation

An improved PE algorithm for continuous nonlinear system identification with application to learning in neural networks

A Sequential Simplex Algorithm For The Continuous Convex Piecewise Linear Network Flow Problem

Adaptive Kernel Leaning Networks with Application to Nonlinear System Identification

System partitioning on MCM using a new neural network model