ParaRNN: An Interpretable and Parallelizable Recurrent Neural Network for Time-Dependent Data

2026-05-04

AI Summary

Traditional recurrent neural networks (RNNs) struggle to meet the demands of statistical modeling due to poor interpretability and low training efficiency. This work proposes a parallelized RNN architecture, termed ParaRNN, which decouples recurrent dynamics into multiple interpretable, small-scale parallel units through an additive structure, integrating nonparametric regression theory with recursive feature extraction. The proposed method achieves predictive performance comparable to standard RNNs while substantially enhancing model interpretability and training speed. Notably, this study establishes the first non-asymptotic prediction error bound for this class of models, providing rigorous theoretical guarantees for their generalization behavior.

Abstract

The proliferation of large-scale and structurally complex data has spurred the integration of machine learning methods into statistical modeling. Recurrent neural networks (RNNs), a foundational class of models for time-dependent data, can be viewed as nonlinear extensions of classical autoregressive moving average models. Despite their flexibility and empirical success in machine learning, RNNs often suffer from limited interpretability and slow training, which hinders their use in statistics. This paper proposes the Parallelized RNN (ParaRNN), a novel model composed of multiple small recurrent units. ParaRNN admits an additive representation that decouples recurrent dynamics into interpretable components, whose behavior can be characterized through recurrence features. This interpretability enables its applications in nonparametric regression for time-dependent data, while the design also allows efficient parallelization. The approximation capacity and non-asymptotic prediction error bounds in a nonparametric regression setting are established for ParaRNN. Empirical results on three sequential modeling tasks further demonstrate that ParaRNN achieves performance comparable to vanilla RNNs while offering improved interpretability and efficiency.

Problem

• interpretability
• training efficiency
• recurrent neural networks
• time-dependent data
• statistical modeling

Innovation

• interpretable RNN
• parallelizable architecture
• additive decomposition
• recurrence features
• nonparametric regression