Deep Residual Learning and PDEs on Manifold

Abstract

In this paper, we formulate the deep residual network (ResNet) as a control problem of transport equation. In ResNet, the transport equation is solved along the characteristics. Based on this observation, deep neural network is closely related to the control problem of PDEs on manifold. We propose several models based on transport equation and Hamilton-Jacobi equation. The discretization of these PDEs on point cloud is also discussed.

References in corpus (1)

• Solving high-dimensional partial differential equations using deep learning

Cited by in corpus (16)

• A Review on Deep Learning in Medical Image Reconstruction
• Beyond Finite Layer Neural Networks: Bridging Deep Architectures and Numerical Differential Equations
• Physics-Informed Supervised Residual Learning for Electromagnetic Modeling
• Stochastic Training of Residual Networks: a Differential Equation Viewpoint
• ResNets Ensemble via the Feynman-Kac Formalism to Improve Natural and Robust Accuracies
• Physics-Aware Neural Implicit Solvers for multiscale, parametric PDEs with applications in heterogeneous media
• Selection dynamics for deep neural networks
• A Flow Model of Neural Networks
• Machine-learning-based methods for output only structural modal identification
• Translating Diffusion, Wavelets, and Regularisation into Residual Networks
• Deep Global Model Reduction Learning in Porous Media Flow Simulation
• Prediction of Discretization of GMsFEM using Deep Learning
• Differential equations as models of deep neural networks
• Machine Learning and Control Theory
• An Integrated Approach to Produce Robust Models with High Efficiency
• Graph Interpolating Activation Improves Both Natural and Robust Accuracies in Data-Efficient Deep Learning