MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers

Data augmentation, generating new data that are similar but not same with the original data by making a series of transformations to the original data, is one of the mainstream methods to alleviate the problem of insufficient data. Instead of augmenting input data, this paper proposes a method for a...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Dongchen Liu, Lun Zhang, Xiansen Jiang, Caixia Su, Yufeng Fan, Yongfeng Cao
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Data augmentation extrapolation feature space Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/434a0ebb58f24674a9e21dac64037db1
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:434a0ebb58f24674a9e21dac64037db1
record_format	dspace
spelling	oai:doaj.org-article:434a0ebb58f24674a9e21dac64037db12021-11-17T00:00:18ZMEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers2169-353610.1109/ACCESS.2021.3125841https://doaj.org/article/434a0ebb58f24674a9e21dac64037db12021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9606654/https://doaj.org/toc/2169-3536Data augmentation, generating new data that are similar but not same with the original data by making a series of transformations to the original data, is one of the mainstream methods to alleviate the problem of insufficient data. Instead of augmenting input data, this paper proposes a method for augmenting features in the middle layers of deep models, called MEML (Mean Extrapolation in Middle Layers). It gets the features outputted by any middle layer of deep models and then generates new features by doing mean extrapolation with some randomly selected features. After that, it replaces these selected features with their corresponding new ones and keeps the labels unchanged, and then lets the new composed output continue to propagate forward. Experiments on two classic deep neural network models and three image datasets show that our MEML method can significantly improve the model classification accuracy and outperform in most experiments the state-of-the-art feature space augmentation methods that factor dropout and K Nearest Neighbors extrapolation. Interestingly, when coupled with some input space augmentation methods, e.g., rotation and horizontal flip, our MEML method further improves the performance of two deep models on three datasets, implying that the input space augmentation methods and MEML could complement each other.Dongchen LiuLun ZhangXiansen JiangCaixia SuYufeng FanYongfeng CaoIEEEarticleData augmentationextrapolationfeature spaceElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 151621-151630 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Data augmentation extrapolation feature space Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Data augmentation extrapolation feature space Electrical engineering. Electronics. Nuclear engineering TK1-9971 Dongchen Liu Lun Zhang Xiansen Jiang Caixia Su Yufeng Fan Yongfeng Cao MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
description	Data augmentation, generating new data that are similar but not same with the original data by making a series of transformations to the original data, is one of the mainstream methods to alleviate the problem of insufficient data. Instead of augmenting input data, this paper proposes a method for augmenting features in the middle layers of deep models, called MEML (Mean Extrapolation in Middle Layers). It gets the features outputted by any middle layer of deep models and then generates new features by doing mean extrapolation with some randomly selected features. After that, it replaces these selected features with their corresponding new ones and keeps the labels unchanged, and then lets the new composed output continue to propagate forward. Experiments on two classic deep neural network models and three image datasets show that our MEML method can significantly improve the model classification accuracy and outperform in most experiments the state-of-the-art feature space augmentation methods that factor dropout and K Nearest Neighbors extrapolation. Interestingly, when coupled with some input space augmentation methods, e.g., rotation and horizontal flip, our MEML method further improves the performance of two deep models on three datasets, implying that the input space augmentation methods and MEML could complement each other.
format	article
author	Dongchen Liu Lun Zhang Xiansen Jiang Caixia Su Yufeng Fan Yongfeng Cao
author_facet	Dongchen Liu Lun Zhang Xiansen Jiang Caixia Su Yufeng Fan Yongfeng Cao
author_sort	Dongchen Liu
title	MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
title_short	MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
title_full	MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
title_fullStr	MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
title_full_unstemmed	MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers
title_sort	meml: a deep data augmentation method by mean extrapolation in middle layers
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/434a0ebb58f24674a9e21dac64037db1
work_keys_str_mv	AT dongchenliu memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers AT lunzhang memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers AT xiansenjiang memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers AT caixiasu memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers AT yufengfan memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers AT yongfengcao memladeepdataaugmentationmethodbymeanextrapolationinmiddlelayers
_version_	1718426044486123520

MEML: A Deep Data Augmentation Method by Mean Extrapolation in Middle Layers

Ejemplares similares