Convolutional neural networks with constrained shortcut connections

Qiyan Du; Tao Zhang

doi:10.1117/12.2656831

20 October 2022 Convolutional neural networks with constrained shortcut connections

Qiyan Du, Tao Zhang

Proceedings Volume 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022); 1245156 (2022) https://doi.org/10.1117/12.2656831
Event: 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 2022, Chongqing, China

Abstract

This paper proposes a novel network architecture to integrate features in the shallow and the deep layers. The architecture is composed of three components, including the shortcut connection, global representation operator, and regularizer. The shortcut connection alleviates the vanishing gradient problem by feeding features from the hidden layer to the classifier in the last layer. The global representation operator aims to reduce the features’ dimension to mitigate the over-fitting problem. The regularizer is designed to relieve the degradation of network depth which might be caused by the shortcut connection. The results on Optical Coherence tomography (OCT) show that the proposed architecture is an effective regularization method that can mitigate the over-fitting problem.

Citation Download Citation

Qiyan Du and Tao Zhang "Convolutional neural networks with constrained shortcut connections", Proc. SPIE 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 1245156 (20 October 2022); https://doi.org/10.1117/12.2656831

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
4 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Network architectures

Neural networks

Convolutional neural networks

Optical coherence tomography

Mechanical engineering

Data modeling

Machine vision

Show All Keywords

Keywords/Phrases

Search In:

Publication Years