2023/08/03 by Yao Liu, Liu, Yao, Hang Shao +3 · 1 citation
Computer Science · Engineering · Mathematics · Physics and Astronomy · #Algorithm #Architecture #Artificial intelligence #Artificial neural network #Class (philosophy) #Computational Physics and Python Applications #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Convolutional neural network #Deep learning #Engineering #FOS: Computer and information sciences #Geography #Geometry #Image (mathematics) #Machine Learning (cs.LG) #Mathematical analysis #Mathematics #Model Reduction and Neural Networks #Network architecture #Neural Networks and Applications #Neural and Evolutionary Computing (cs.NE) #Partial differential equation #Pattern recognition (psychology) #Perspective (graphical) #Property (philosophy) #Symmetry (geometry) #Task (project management)
paper · pdf · doi:10.48550/arxiv.2308.01621
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2023/08/03 · openalex created_date 2023/08/19 · openalex updated_date 2026/07/28
This paper introduces a new Convolutional Neural Network (ConvNet) architecture inspired by a class of partial differential equations (PDEs) called quasi-linear hyperbolic systems. With comparable performance on the image classification task, it allows for the modification of the weights via a continuous group of symmetry. This is a significant shift from traditional models where the architecture and weights are essentially fixed. We wish to promote the (internal) symmetry as a new desirable property for a neural network, and to draw attention to the PDE perspective in analyzing and interpreting ConvNets in the broader Deep Learning community.