Chebyshev Feature Neural Network for Accurate Function Approximation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Zhongshu, Chen, Yuan, Xiu, Dongbin
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915074770206720
author Xu, Zhongshu
Chen, Yuan
Xiu, Dongbin
author_facet Xu, Zhongshu
Chen, Yuan
Xiu, Dongbin
contents We present a new Deep Neural Network (DNN) architecture capable of approximating functions up to machine accuracy. Termed Chebyshev Feature Neural Network (CFNN), the new structure employs Chebyshev functions with learnable frequencies as the first hidden layer, followed by the standard fully connected hidden layers. The learnable frequencies of the Chebyshev layer are initialized with exponential distributions to cover a wide range of frequencies. Combined with a multi-stage training strategy, we demonstrate that this CFNN structure can achieve machine accuracy during training. A comprehensive set of numerical examples for dimensions up to $20$ are provided to demonstrate the effectiveness and scalability of the method.
format Preprint
id arxiv_https___arxiv_org_abs_2409_19135
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Chebyshev Feature Neural Network for Accurate Function Approximation
Xu, Zhongshu
Chen, Yuan
Xiu, Dongbin
Machine Learning
Numerical Analysis
Neural and Evolutionary Computing
65T40, 68T01
We present a new Deep Neural Network (DNN) architecture capable of approximating functions up to machine accuracy. Termed Chebyshev Feature Neural Network (CFNN), the new structure employs Chebyshev functions with learnable frequencies as the first hidden layer, followed by the standard fully connected hidden layers. The learnable frequencies of the Chebyshev layer are initialized with exponential distributions to cover a wide range of frequencies. Combined with a multi-stage training strategy, we demonstrate that this CFNN structure can achieve machine accuracy during training. A comprehensive set of numerical examples for dimensions up to $20$ are provided to demonstrate the effectiveness and scalability of the method.
title Chebyshev Feature Neural Network for Accurate Function Approximation
topic Machine Learning
Numerical Analysis
Neural and Evolutionary Computing
65T40, 68T01
url https://arxiv.org/abs/2409.19135