What question did this study set out to answer?

To develop a hybrid network combining BiLSTM and transformer architectures for effective forecasting of chaotic dynamics and nonlinear observability.

March 4, 2026AIP AdvancesOpen Access

Parallel BiLSTM–transformer networks for forecasting chaotic dynamics

View Full Paper

Ask AI

Bookmark

Authors

Hybrid architecture enhances predictive accuracy in forecasting chaotic time series and nonlinear observability.

To develop a hybrid network combining BiLSTM and transformer architectures for effective forecasting of chaotic dynamics and nonlinear observability.
Developed a parallel architecture integrating BiLSTM and transformer networks.
Utilized time-delay embeddings for autonomous evolution prediction.
Fused representations from both branches using a feature-integration module.
Evaluated on the Lorenz system across two distinct forecasting tasks.
Hybrid model outperforms single-branch architectures in both forecasting tasks.
Achieved stable reconstruction of unmeasured variables over long time horizons.
Improved predictive accuracy and robustness for chaotic time series forecasting.

Ma et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark