Efficient ensemble for image-based identification of Pneumonia utilizing deep CNN and SGD with warm restarts
作者：
Highlights：
• A new ensemble method SGDRE utilizing CNN and SGD with warm restarts is proposed.
• The SGDRE addresses the issues of generalization, dataset size and time complexity.
• Using SGDRE an ensemble of CNN models is built without increasing training time.
• The SGDRE efficiently and effectively identifies pneumonia from chest X-ray images.
• The SGDRE on average achieves an accuracy of 96.26% and outperforms compared methods.
摘要
•A new ensemble method SGDRE utilizing CNN and SGD with warm restarts is proposed.•The SGDRE addresses the issues of generalization, dataset size and time complexity.•Using SGDRE an ensemble of CNN models is built without increasing training time.•The SGDRE efficiently and effectively identifies pneumonia from chest X-ray images.•The SGDRE on average achieves an accuracy of 96.26% and outperforms compared methods.
论文关键词：Deep learning,Convolutional neural network,Ensemble model,Image classification,Stochastic gradient descent,Identification of pneumonia
论文评审过程：Received 2 April 2020, Revised 10 August 2021, Accepted 29 August 2021, Available online 7 September 2021, Version of Record 15 September 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115834
