Stabilizing Convolutional Neural Networks with Modified Runge–Kutta Integration

DOI

10.2991/978-94-6239-713-2_46How to use a DOI?

Keywords

Convolutional Neural Networks; Modified Runge–Kutta; Batch Normalization; CIFAR-10

Abstract

Convolutional Neural Networks (CNNs) are commonly employed for image classification, but training stability and generalization are difficult due to discretization errors during feature propagation. In this paper, we provide a Modified Runge-Kutta (MRK4) integration approach as a parameter-free post-pooling stabilizer for typical CNN architectures. Experiments on the CIFAR-10 dataset compare a baseline CNN, a CNN with Batch Normalization (BN), and MRK4-augmented variants across different random seeds. Results show that the proposed CNN-BN-MRK4 model increases mean classification accuracy by up to 3.5% and reduces validation loss by approximately 18% compared to CNN-BN, while keeping comparable inference time. Furthermore, the MRK4-based technique exhibits lower performance variance and higher generalization stability. These results show that higher-order numerical integration can effectively stabilize deep CNN training without introducing new learnable parameters.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Sidhartha Sankar Pradhan
AU  - Subhendu Sekhar Sahoo
PY  - 2026
DA  - 2026/06/25
TI  - Stabilizing Convolutional Neural Networks with Modified Runge–Kutta Integration
BT  - Proceedings of the International Conference on Advances in Computing Technology and Artificial Intelligence (COMPUTATIA 2026)
PB  - Atlantis Press
SP  - 623
EP  - 633
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-713-2_46
DO  - 10.2991/978-94-6239-713-2_46
ID  - Pradhan2026
ER  -