Research map / Time series / Long-term forecasting
DEPTS: Deep Expansion Learning for Periodic Time Series Forecasting
Wei Fan, Shun Zheng, Xiaohan Yi, Wei Cao, Yanjie Fu, Jiang Bian, Tie-Yan Liu
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 |
|---|---|---|---|
| Model periodicity explicitly with DEPTS | PTS forecasting · Electricity (2014-12-25) | nd: 0.171 -> 0.139 | Improved |
| Model periodicity explicitly with DEPTS | PTS forecasting · Caiso (2020-10-01) | nd: 0.026 -> 0.020 | Improved |
| Add local-block inputs with periodic effects subtracted (ablation) | DEPTS · Electricity (2014-12-25), J=4 | nd: 0.15870 -> 0.14931 | Improved |
| Add periodic-block contributions to cumulative forecasts (ablation) | DEPTS · M4 (Hourly) | nd: 0.02161 -> 0.02050 | Improved |
| Add residual updates of periodic states between layers (ablation) | DEPTS · M4 (Hourly) | nd: 0.02252 -> 0.02050 | Improved |
| Add periodic blocks rather than omit explicit periodic modeling (ablation) | DEPTS · Electricity (2014-12-25), J=4 | nd: 0.20615 -> 0.14931 | Improved |
| Initialize periodic coefficients from data rather than randomly (ablation) | DEPTS · Electricity (2014-12-25), J=4 | nd: 0.17677 -> 0.14931 | Improved |
| Tune initialized periodic coefficients end-to-end rather than fix them (ablation) | DEPTS · Electricity (2014-12-25), J=4 | nd: 0.16756 -> 0.14931 | Improved |
| Use separate local and periodic processing instead of feeding periodic state as a second covariate channel (ablation) | DEPTS · Electricity (2014-12-25), J=4 | nd: 0.15743 -> 0.14931 | Improved |
| Increase the maximum selected periods J (ablation) | DEPTS · Electricity (2014-12-25) | nd: 0.14931 -> 0.13915 | Improved |
| Increase the maximum selected periods J (ablation) | DEPTS · Traffic (2009-03-24) | nd: 0.10745 -> 0.11076 | Mixed |
Later papers that compared with this method
1 later accepted paper reports Depts as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Non-autoregressive Conditional Diffusion Models for Time Series Prediction (ICML 2023 · better on 0 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SynTSBench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series (NeurIPS 2025 · similarity 0.86)
- Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization (ICML 2025 · similarity 0.86)
- CycleNet: Enhancing Time Series Forecasting through Modeling Periodic Patterns (NeurIPS 2024 · similarity 0.86)
- Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024 · similarity 0.85)
- TimeBase: The Power of Minimalism in Efficient Long-term Time Series Forecasting (ICML 2025 · similarity 0.85)
- TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024 · similarity 0.85)
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.
