Research map / Time series / Long-term forecasting
Correlation Recurrent Units: A Novel Neural Architecture for Improving the Predictive Performance of Time-Series Data
Sunghyun Sim, Dohee Kim, Hyerim Bae
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the STL cell inside the recurrent cell | RNN · UTD1 US GDP, 1-step | RMSE: 120.707 -> 110.712 | Improved |
| add the STL cell inside the cell instead of STL decomposition across separate networks | RNN · UTD1 US GDP, 1-step | RMSE: 112.889 -> 110.712 | Improved |
| apply network-level STL decomposition with separate RNNs | RNN · UTD1 US GDP, 1-step | RMSE: 120.707 -> 112.889 | Improved |
| add autocorrelation and correlation gates to the STL cell (ablation) | RNN-STLC · UTD1 US GDP, 1-step | RMSE: 110.712 -> 104.766 | Improved |
| add autocorrelation and correlation gates to the STL cell (ablation) | RNN-STLC · MTD3 world stock indices, 1-step | RMSE: 107.55 -> 95.17 | Improved |
| add autocorrelation and correlation gates to the STL cell (ablation) | RNN-STLC · MTD1 big-tech stocks, 1-step | RMSE: 21.28 -> 20.15 | Improved |
| use STL cell instead of STL decomposition | LSTM-STLD · MTD3 world stock indices, 1-step | RMSE: 184.89 -> 115.91 | Improved |
| replace DSANet with CRU | DSANet · MTD1 big-tech stocks, 1-step | RMSE: 28.32 -> 20.15 | Improved |
| replace DSANet with CRU | DSANet · MTD2 Baltic dry indices, 6-step | RMSE: 184.18 -> 167.26 | Improved |
| replace DSANet with CRU | DSANet · MTD3 world stock indices, 12-step | RMSE: 432.70 -> 499.03 | Hurt |
| replace DSANet with CRU | DSANet · MTD1 big-tech stocks, 12-step | RMSE: 95.27 -> 96.49 | Hurt |
| replace LSTNet-A with CRU | LSTNet-A · MTD2 Baltic dry indices, 12-step | RMSE: 285.21 -> 270.16 | Improved |
Closest papers by meaning
- Unlocking the Power of LSTM for Long Term Time Series Forecasting (AAAI 2025 · similarity 0.86)
- TStruct: Learning Shared Temporal Structures for Long-Term Time-Series Forecasting (NeurIPS 2026 · similarity 0.85)
- Efficient and Effective Time-Series Forecasting with Spiking Neural Networks (ICML 2024 · similarity 0.85)
- TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting (NeurIPS 2025 · similarity 0.85)
- Temporal Query Network for Efficient Multivariate Time Series Forecasting (ICML 2025 · similarity 0.84)
- PGN: The RNN's New Successor is Effective for Long-Range Time Series Forecasting (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
