Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination
Designing a Convolutional Neural Networks (CNN) is a complex task and requires expert knowledge to optimize the performance and network architecture. In this paper, a novel data-driven approach is proposed to determine the architecture of CNN models. The proposed Autonomous Convolutional Neural Netw...
Saved in:
Main Authors: | , , , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2022
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/163879 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-163879 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1638792022-12-21T02:36:18Z Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination Aradhya, Abhay M S Ashfahani, Andri Angelina, Fienny Pratama, Mahardhika de Mello, Rodrigo Fernandes Sundaram, Suresh School of Computer Science and Engineering Engineering::Computer science and engineering Evolving Intelligent Systems Convolution Neural Networks Designing a Convolutional Neural Networks (CNN) is a complex task and requires expert knowledge to optimize the performance and network architecture. In this paper, a novel data-driven approach is proposed to determine the architecture of CNN models. The proposed Autonomous Convolutional Neural Networks (AutoCNNThe executable code and original numerical results can be downloaded from (https://tinyurl.com/AutoCNN)) algorithm introduces data driven strategies for addition of new convolutional layers, pruning of redundant filters and training cycle optimization. AutoCNN is evaluated using MNIST, MNIST-rot-back-image, Fashion MNIST and the ADHD200 datasets to measure the performance on small datasets with varied feature distributions. The results indicate that AutoCNN optimizes the CNN network architecture and helps maximise the classification performance. The data-driven network determination approach introduced in this paper was found to not only provides competitive performance similar to existing evolutionary computation based network determination algorithms in literature, but was found to be an effective optimization tool to improve the performance of existing CNN architectures. Further, the AutoCNN was found to highly immune to noise in the dataset and has proven to be effective method to transfer knowledge between related datasets. Therefore, the AutoCNN is a highly versatile CNN architecture determination tool that has a wide range of applications in the field of autonomous driving, medical image analysis, image enhancement, camera based security monitoring and image based fault detection. Ministry of Education (MOE) This work is supported by Ministry of Education, Republic of Singapore, Tier 1 Research Grant (RG90/20). 2022-12-21T02:36:18Z 2022-12-21T02:36:18Z 2022 Journal Article Aradhya, A. M. S., Ashfahani, A., Angelina, F., Pratama, M., de Mello, R. F. & Sundaram, S. (2022). Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination. Information Sciences, 607, 638-653. https://dx.doi.org/10.1016/j.ins.2022.05.100 0020-0255 https://hdl.handle.net/10356/163879 10.1016/j.ins.2022.05.100 2-s2.0-85131828621 607 638 653 en RG90/20 Information Sciences © 2022 Elsevier Inc. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Computer science and engineering Evolving Intelligent Systems Convolution Neural Networks |
spellingShingle |
Engineering::Computer science and engineering Evolving Intelligent Systems Convolution Neural Networks Aradhya, Abhay M S Ashfahani, Andri Angelina, Fienny Pratama, Mahardhika de Mello, Rodrigo Fernandes Sundaram, Suresh Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
description |
Designing a Convolutional Neural Networks (CNN) is a complex task and requires expert knowledge to optimize the performance and network architecture. In this paper, a novel data-driven approach is proposed to determine the architecture of CNN models. The proposed Autonomous Convolutional Neural Networks (AutoCNNThe executable code and original numerical results can be downloaded from (https://tinyurl.com/AutoCNN)) algorithm introduces data driven strategies for addition of new convolutional layers, pruning of redundant filters and training cycle optimization. AutoCNN is evaluated using MNIST, MNIST-rot-back-image, Fashion MNIST and the ADHD200 datasets to measure the performance on small datasets with varied feature distributions. The results indicate that AutoCNN optimizes the CNN network architecture and helps maximise the classification performance. The data-driven network determination approach introduced in this paper was found to not only provides competitive performance similar to existing evolutionary computation based network determination algorithms in literature, but was found to be an effective optimization tool to improve the performance of existing CNN architectures. Further, the AutoCNN was found to highly immune to noise in the dataset and has proven to be effective method to transfer knowledge between related datasets. Therefore, the AutoCNN is a highly versatile CNN architecture determination tool that has a wide range of applications in the field of autonomous driving, medical image analysis, image enhancement, camera based security monitoring and image based fault detection. |
author2 |
School of Computer Science and Engineering |
author_facet |
School of Computer Science and Engineering Aradhya, Abhay M S Ashfahani, Andri Angelina, Fienny Pratama, Mahardhika de Mello, Rodrigo Fernandes Sundaram, Suresh |
format |
Article |
author |
Aradhya, Abhay M S Ashfahani, Andri Angelina, Fienny Pratama, Mahardhika de Mello, Rodrigo Fernandes Sundaram, Suresh |
author_sort |
Aradhya, Abhay M S |
title |
Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
title_short |
Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
title_full |
Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
title_fullStr |
Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
title_full_unstemmed |
Autonomous CNN (AutoCNN): a data-driven approach to network architecture determination |
title_sort |
autonomous cnn (autocnn): a data-driven approach to network architecture determination |
publishDate |
2022 |
url |
https://hdl.handle.net/10356/163879 |
_version_ |
1753801145811206144 |