Stabilized neural ordinary differential equation for text classification in natural language processing

doi:10.1016/j.neucom.2025.132008

Stabilized neural ordinary differential equation for text classification in natural language processing

2025-11-03

0

PRE

AI

DOI：10.1016/j.neucom.2025.132008

Original

Original request for help

Save

Abstract

En 中文

Neural ordinary differential equations (ODEs) are widely used in the field of deep learning. This paper proposes an innovative neural ODE layer structure, which is a two-layer neural network with a time delay. We prove the existence and uniqueness of the equilibrium point in the proposed neural ODE layer using Brouwer degree theory and the concept of Lipschitz continuity. In addition, a set of stability criteria is derived to ensure the global asymptotic stability (GAS) of the neural ODE layer. Then, we incorporate stability constraints as regularizers during model training and evaluate their performance on a text classification task in natural language processing (NLP). The experimental results demonstrate that the proposed neural ODE layer achieves faster convergence and higher accuracy compared with the baseline model, the long short-term memory network (LSTM). The research results fully demonstrate that the proposed method has broad application prospects in deep learning.

Stabilized neural ordinary differential equation for text classification in natural language processing

Abstract

Journal

Organization