Towards Understanding Normalization in Neural ODEs

TL;DR

This paper explores the impact of normalization techniques on neural ODEs, achieving a new high accuracy of 93% on CIFAR-10, and enhances understanding of their role in deep learning models.

Contribution

It provides the first comprehensive analysis of normalization effects in neural ODEs and reports the highest CIFAR-10 accuracy for such models.

Findings

Normalization significantly influences neural ODE performance

Achieved 93% accuracy on CIFAR-10 with neural ODEs

Provides insights into normalization's role in deep learning

Abstract

Normalization is an important and vastly investigated technique in deep learning. However, its role for Ordinary Differential Equation based networks (neural ODEs) is still poorly understood. This paper investigates how different normalization techniques affect the performance of neural ODEs. Particularly, we show that it is possible to achieve 93% accuracy in the CIFAR-10 classification task, and to the best of our knowledge, this is the highest reported accuracy among neural ODEs tested on this problem.