A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification

Yun Zhang; Qinglong Hua; Dan Xu; Hongbo Li; Hui Lin Mu

doi:10.1109/RADAR41533.2019.171298

A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification

Yun Zhang
, Qinglong Hua
, Dan Xu
, Hongbo Li
, Hui Lin Mu

School of Electronics and Information Engineering, Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

7 Scopus citations

Abstract

It is well known that the activation function and the gradient descent optimization algorithm have a great influence on convolutional neural network (CNN). Based on the Adam optimization algorithm, this paper proposes a CvAdam optimization algorithm suitable for complex-valued convolutional neural network (CV-CNN), and then in the typical polarization SAR image classification task, Adam and CvAdam were compared using four different activation functions of sigmoid, tanh, Leakey-ReLU and ELU. Experiments on the benchmark dataset of Oberpfaffenhofen show that CvAdam performs better than Adam in both convergence speed and accuracy, no matter which activation function is used.

Original language	English
Title of host publication	2019 International Radar Conference, RADAR 2019
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781728126609
DOIs	https://doi.org/10.1109/RADAR41533.2019.171298
State	Published - Sep 2019
Externally published	Yes
Event	2019 International Radar Conference, RADAR 2019 - Toulon, France Duration: 23 Sep 2019 → 27 Sep 2019

Publication series

Name	2019 International Radar Conference, RADAR 2019

Conference

Conference	2019 International Radar Conference, RADAR 2019
Country/Territory	France
City	Toulon
Period	23/09/19 → 27/09/19

Keywords

Complex-valued convolutional neural network (CV-CNN)
activation function
deep learning
optimization algorithm
synthetic aperture radar (SAR)
terrain classification

Access to Document

10.1109/RADAR41533.2019.171298

Cite this

Zhang, Y., Hua, Q., Xu, D., Li, H., & Mu, H. L. (2019). A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification. In 2019 International Radar Conference, RADAR 2019 Article 9079055 (2019 International Radar Conference, RADAR 2019). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/RADAR41533.2019.171298

@inproceedings{9608b3635d6b403eab6d839b69e76ddd,

title = "A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification",

abstract = "It is well known that the activation function and the gradient descent optimization algorithm have a great influence on convolutional neural network (CNN). Based on the Adam optimization algorithm, this paper proposes a CvAdam optimization algorithm suitable for complex-valued convolutional neural network (CV-CNN), and then in the typical polarization SAR image classification task, Adam and CvAdam were compared using four different activation functions of sigmoid, tanh, Leakey-ReLU and ELU. Experiments on the benchmark dataset of Oberpfaffenhofen show that CvAdam performs better than Adam in both convergence speed and accuracy, no matter which activation function is used.",

keywords = "Complex-valued convolutional neural network (CV-CNN), activation function, deep learning, optimization algorithm, synthetic aperture radar (SAR), terrain classification",

author = "Yun Zhang and Qinglong Hua and Dan Xu and Hongbo Li and Mu, \{Hui Lin\}",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 2019 International Radar Conference, RADAR 2019 ; Conference date: 23-09-2019 Through 27-09-2019",

year = "2019",

month = sep,

doi = "10.1109/RADAR41533.2019.171298",

language = "英语",

series = "2019 International Radar Conference, RADAR 2019",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2019 International Radar Conference, RADAR 2019",

address = "美国",

}

Zhang, Y , Hua, Q, Xu, D, Li, H & Mu, HL 2019, A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification. in 2019 International Radar Conference, RADAR 2019., 9079055, 2019 International Radar Conference, RADAR 2019, Institute of Electrical and Electronics Engineers Inc., 2019 International Radar Conference, RADAR 2019, Toulon, France, 23/09/19. https://doi.org/10.1109/RADAR41533.2019.171298

A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification. / Zhang, Yun ; Hua, Qinglong; Xu, Dan et al.
2019 International Radar Conference, RADAR 2019. Institute of Electrical and Electronics Engineers Inc., 2019. 9079055 (2019 International Radar Conference, RADAR 2019).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification

AU - Zhang, Yun

AU - Hua, Qinglong

AU - Xu, Dan

AU - Li, Hongbo

AU - Mu, Hui Lin

PY - 2019/9

Y1 - 2019/9

N2 - It is well known that the activation function and the gradient descent optimization algorithm have a great influence on convolutional neural network (CNN). Based on the Adam optimization algorithm, this paper proposes a CvAdam optimization algorithm suitable for complex-valued convolutional neural network (CV-CNN), and then in the typical polarization SAR image classification task, Adam and CvAdam were compared using four different activation functions of sigmoid, tanh, Leakey-ReLU and ELU. Experiments on the benchmark dataset of Oberpfaffenhofen show that CvAdam performs better than Adam in both convergence speed and accuracy, no matter which activation function is used.

AB - It is well known that the activation function and the gradient descent optimization algorithm have a great influence on convolutional neural network (CNN). Based on the Adam optimization algorithm, this paper proposes a CvAdam optimization algorithm suitable for complex-valued convolutional neural network (CV-CNN), and then in the typical polarization SAR image classification task, Adam and CvAdam were compared using four different activation functions of sigmoid, tanh, Leakey-ReLU and ELU. Experiments on the benchmark dataset of Oberpfaffenhofen show that CvAdam performs better than Adam in both convergence speed and accuracy, no matter which activation function is used.

KW - Complex-valued convolutional neural network (CV-CNN)

KW - activation function

KW - deep learning

KW - optimization algorithm

KW - synthetic aperture radar (SAR)

KW - terrain classification

UR - https://www.scopus.com/pages/publications/85084955737

U2 - 10.1109/RADAR41533.2019.171298

DO - 10.1109/RADAR41533.2019.171298

M3 - 会议稿件

AN - SCOPUS:85084955737

T3 - 2019 International Radar Conference, RADAR 2019

BT - 2019 International Radar Conference, RADAR 2019

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2019 International Radar Conference, RADAR 2019

Y2 - 23 September 2019 through 27 September 2019

ER -

A Complex-Valued Convolutional Neural Network with Different Activation Functions in Polarimetric SAR Image Classification

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this