ePrints

Enhanced Pooling Method for Convolutional Neural Networks based on Optimal Search Theory

Lookup NU author(s): Xin Lai, Zeyu Fu, Dr Mohsen Naqvi, Professor Jonathon Chambers

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.

Abstract

To obtain the best pooling effect and higher accuracy in image recognition, an improved method based on optimalsearch theory for the pooling layer of convolutional neural networks (CNNs) is proposed. The purpose is to solve the problemsof the traditional pooling method, namely that it is too simplistic and it is difficult to extract effective features. The basic principleand network structure of a CNN are introduced in the paper. A new optimum-pooling method is proposed, and we study how toobtain the maximum probability to detect the target function under the constrained condition. Comparison experiments of differentpooling methods are performed on three widely used datasets: LFW, CIFAR-10, and ImageNet. The experimental results showthat the proposed method has the characteristics of more effective feature extraction and wide adaptability, and leads to higheraccuracy and lower error rate in image recognition.

Publication metadata

Author(s): Lai X, Zhou L, Fu Z, Naqvi SM, Chambers J

Publication type: Article

Publication status: Published

Journal: IET Image Processing

Year: 2019

Volume: 13

Issue: 12

Pages: 2152-2161

Print publication date: 01/10/2019

Online publication date: 01/08/2019

Acceptance date: 16/07/2019

ISSN (print): 1751-9659

ISSN (electronic): 1751-9667

Publisher: IET

URL: https://doi.org/10.1049/iet-ipr.2018.6322

DOI: 10.1049/iet-ipr.2018.6322

Altmetrics

Share