2019/11/23 by Juncai He, He, Juncai, Yuyan Chen +5 · 2 citations
Computer Science · Engineering · Mathematics · #Algorithm #Artificial intelligence #Computer science #Convolutional neural network #Data mining #Feature (linguistics) #Image (mathematics) #Medical Image Segmentation Techniques #Neural Networks and Applications #Pattern recognition (psychology) #Residual neural network #Sparse and Compressive Sensing Techniques #cs.CV #cs.NA #eess.IV #math.NA
paper · pdf · doi:10.48550/arxiv.1911.10428
published in arXiv (Cornell University) (Cornell University) · 15 page, 2 figures
openalex publication_date 2019/11/23 · arxiv created 2020/07/06 · arxiv updated 2020/07/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In this paper, we propose a constrained linear data-feature mapping model as an interpretable mathematical model for image classification using convolutional neural network (CNN) such as the ResNet. From this viewpoint, we establish the detailed connections in a technical level between the traditional iterative schemes for constrained linear system and the architecture for the basic blocks of ResNet. Under these connections, we propose some natural modifications of ResNet type models which will have less parameters but still maintain almost the same accuracy as these corresponding original models. Some numerical experiments are shown to demonstrate the validity of this constrained learning data-feature mapping assumption.