A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method

In response to the challenge of overfitting, which may lead to a decline in network generalization performance, this paper proposes a new regularization technique, called the class-based decorrelation method (CDM). Specifically, this method views the neurons in a specific hidden layer as base learne...

Full description

Bibliographic Details
Main Authors: Chenguang Zhang, Tian Liu, Xuejiao Du
Format: Article
Language:English
Published: MDPI AG 2023-12-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/26/1/7
_version_ 1827372373646508032
author Chenguang Zhang
Tian Liu
Xuejiao Du
author_facet Chenguang Zhang
Tian Liu
Xuejiao Du
author_sort Chenguang Zhang
collection DOAJ
description In response to the challenge of overfitting, which may lead to a decline in network generalization performance, this paper proposes a new regularization technique, called the class-based decorrelation method (CDM). Specifically, this method views the neurons in a specific hidden layer as base learners, and aims to boost network generalization as well as model accuracy by minimizing the correlation among individual base learners while simultaneously maximizing their class-conditional correlation. Intuitively, CDM not only promotes diversity among the hidden neurons, but also enhances their cohesiveness among them when processing samples from the same class. Comparative experiments conducted on various datasets using deep models demonstrate that CDM effectively reduces overfitting and improves classification performance.
first_indexed 2024-03-08T10:57:42Z
format Article
id doaj.art-dec702a9baa24eb4ac128ba3f64f2570
institution Directory Open Access Journal
issn 1099-4300
language English
last_indexed 2024-03-08T10:57:42Z
publishDate 2023-12-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj.art-dec702a9baa24eb4ac128ba3f64f25702024-01-26T16:22:44ZengMDPI AGEntropy1099-43002023-12-01261710.3390/e26010007A Deep Neural Network Regularization Measure: The Class-Based Decorrelation MethodChenguang Zhang0Tian Liu1Xuejiao Du2School of Mathematics and Statistics, Hainan University, Haikou 570100, ChinaSchool of Information and Communication Engineering, Hainan University, Haikou 570100, ChinaSchool of Mathematics and Statistics, Hainan University, Haikou 570100, ChinaIn response to the challenge of overfitting, which may lead to a decline in network generalization performance, this paper proposes a new regularization technique, called the class-based decorrelation method (CDM). Specifically, this method views the neurons in a specific hidden layer as base learners, and aims to boost network generalization as well as model accuracy by minimizing the correlation among individual base learners while simultaneously maximizing their class-conditional correlation. Intuitively, CDM not only promotes diversity among the hidden neurons, but also enhances their cohesiveness among them when processing samples from the same class. Comparative experiments conducted on various datasets using deep models demonstrate that CDM effectively reduces overfitting and improves classification performance.https://www.mdpi.com/1099-4300/26/1/7deep neural networkgeneralization abilityregularization method
spellingShingle Chenguang Zhang
Tian Liu
Xuejiao Du
A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
Entropy
deep neural network
generalization ability
regularization method
title A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
title_full A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
title_fullStr A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
title_full_unstemmed A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
title_short A Deep Neural Network Regularization Measure: The Class-Based Decorrelation Method
title_sort deep neural network regularization measure the class based decorrelation method
topic deep neural network
generalization ability
regularization method
url https://www.mdpi.com/1099-4300/26/1/7
work_keys_str_mv AT chenguangzhang adeepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod
AT tianliu adeepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod
AT xuejiaodu adeepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod
AT chenguangzhang deepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod
AT tianliu deepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod
AT xuejiaodu deepneuralnetworkregularizationmeasuretheclassbaseddecorrelationmethod