Partial Differential Equation-Regularized Neural Networks: An Application to Image Classification

Jungeun Kim, Seunghyun Hwang, Jeehyun Hwang, Kookjin Lee, Dongeun Lee, Noseong Park

Keywords: partial differential equations, image classification, physics-informed neural networks

TL;DR: Learn a PDE (approximated by a neural network) from data for image classification

Abstract: Differential equations can be used to design neural networks. For instance, neural ordinary differential equations (neural ODEs) can be considered as a continuous generalization of residual networks. In this work, we present a novel partial differential equation (PDE)-based approach for image classification, where we construct a continuous-depth and continuous-width neural network as a form of solutions of PDEs, and the PDEs defining the evolution of the solutions also are learned from data. Owing to the recent advancement of identifying PDEs, the presented novel concept, called PR-Net, can be implemented. Our method shows comparable (or better) accuracy and robustness for various datasets and tasks in comparison with neural ODEs and Isometric MobileNet V3. Thanks to the efficient nature of PR-Net, it is suitable to be deployed in resource-scarce environments, e.g., deployed instead of MobileNet.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning