Long-term series forecasting with Query Selector -- efficient model of sparse attention
About
Various modifications of TRANSFORMER were recently used to solve time-series forecasting problem. We propose Query Selector - an efficient, deterministic algorithm for sparse attention matrix. Experiments show it achieves state-of-the art results on ETT, Helpdesk and BPI'12 datasets.
Jacek Klimek, Jakub Klimek, Witold Kraskiewicz, Mateusz Topolewski• 2021
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multivariate Time-series Forecasting | ETTh2 (test) | MSE0.4124 | 187 | |
| Multivariate Time-series Forecasting | ETTh1 (test) | MSE0.4226 | 150 | |
| Multivariate Time-series Forecasting | ETTm1 (test) | MSE0.3351 | 83 | |
| Univariate Time Series Forecasting | ETTh1 (test) | MSE0.0436 | 39 | |
| Univariate Time Series Forecasting | ETTm1 (test) | MSE0.0139 | 15 | |
| Activity Prediction | Helpdesk (test) | Accuracy74.3 | 2 | |
| Activity Prediction | BPI'12 (test) | Accuracy79 | 2 |
Showing 7 of 7 rows