Orthogonal Transforms in Neural Networks Amount to Effective Regularization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zając, Krzysztof, Sopot, Wojciech, Wachel, Paweł
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913662033199104
author Zając, Krzysztof
Sopot, Wojciech
Wachel, Paweł
author_facet Zając, Krzysztof
Sopot, Wojciech
Wachel, Paweł
contents We consider applications of neural networks in nonlinear system identification and formulate a hypothesis that adjusting general network structure by incorporating frequency information or other known orthogonal transform, should result in an efficient neural network retaining its universal properties. We show that such a structure is a universal approximator and that using any orthogonal transform in a proposed way implies regularization during training by adjusting the learning rate of each parameter individually. We empirically show in particular, that such a structure, using the Fourier transform, outperforms equivalent models without orthogonality support.
format Preprint
id arxiv_https___arxiv_org_abs_2305_06344
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Orthogonal Transforms in Neural Networks Amount to Effective Regularization
Zając, Krzysztof
Sopot, Wojciech
Wachel, Paweł
Machine Learning
Neural and Evolutionary Computing
Systems and Control
We consider applications of neural networks in nonlinear system identification and formulate a hypothesis that adjusting general network structure by incorporating frequency information or other known orthogonal transform, should result in an efficient neural network retaining its universal properties. We show that such a structure is a universal approximator and that using any orthogonal transform in a proposed way implies regularization during training by adjusting the learning rate of each parameter individually. We empirically show in particular, that such a structure, using the Fourier transform, outperforms equivalent models without orthogonality support.
title Orthogonal Transforms in Neural Networks Amount to Effective Regularization
topic Machine Learning
Neural and Evolutionary Computing
Systems and Control
url https://arxiv.org/abs/2305.06344