Constrained Linear Data-feature Mapping for Image Classification

2019/11/23 by Juncai He, Yuyan Chen, He, Juncai +3
Computer Science · Engineering · #Neural Networks and Applications #Medical Image Segmentation Techniques #Sparse and Compressive Sensing Techniques

paper · pdf · doi:10.48550/arxiv.1911.10428

Abstract

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.

Citations

• Gradient-based learning applied to document recognition
• ImageNet classification with deep convolutional neural networks

Related

• An Interpretive Constrained Linear Model for ResNet and MgNet
• RegNet: Self-Regulated Network for Image Classification
• ResMLP: Feedforward networks for image classification with data-efficient training
• Bag of Tricks for Image Classification with Convolutional Neural Networks
• LeanResNet: A Low-cost Yet Effective Convolutional Residual Networks