Research map / Time series / Long-term forecasting
PHAT: Modeling Period Heterogeneity for Multivariate Time Series Forecasting
Jiaming Ma, Qihe Huang, Haofeng Ma, Guanjun Wang, Sheng Huang, Zhengyang Zhou, Pengkun Wang, Xu Wang, Binwu Wang, Yang Wang
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 period-based variate bucketing (ablation) | PHAT · 12 forecasting datasets | MSE | Improved |
| add the period-aligned attention (PAA) branch (ablation) | PHAT · NN5 and CzeLan | MSE | Improved |
| add the period-offset attention (POA) branch (ablation) | PHAT · 12 forecasting datasets | MSE | Improved |
| add self-attention instead of a single feed-forward net (ablation) | PHAT · 12 forecasting datasets | MSE | Improved |
| model positive and negative dependencies with separate modulators (ablation) | PHAT · periodic positive-negative attention ablation | MSE | Improved |
| use more PNA layers (ablation) | PHAT · hyperparameter sensitivity | MSE | No clear effect |
| use more selected periods K (ablation) | PHAT · datasets with simple periodicity | MSE | Hurt |
| replace best baselines with PHAT | TimeKAN · NYSE | MSE: up to 23.33% | Improved |
| replace TimeKAN with PHAT, fewer parameters | TimeKAN · ETTm1 (horizon 96) | MAE: 0.346 -> 0.330 | Improved |
| replace TimeKAN with PHAT | TimeKAN · Complexity comparison | # parameters: reduces parameter count by up to 91.27% | Improved |
| replace PDF with PHAT on AQShunyi | PDF · AQShunyi MSE (horizon 96) | MSE: 0.647 -> 0.665 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 5 on most shared results.
- TimeKAN — TimeKAN: KAN-based Frequency Decomposition Learning Architecture for Long-term Time Series Forecasting (ICLR 2025) · better on 7 of 8 shared results
- xPatch — xPatch: Dual-Stream Time Series Forecasting with Exponential Seasonal-Trend Decomposition (AAAI 2025) · better on 2 of 2 shared results
- TimeMixer — TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024) · better on 1 of 2 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 1 of 1 shared result
- Pathformer — Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting (ICLR 2024) · better on 1 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
- HDT: Hierarchical Discrete Transformer for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.88)
- FilterTS: Comprehensive Frequency Filtering for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.87)
- PAMod: Modeling Cyclical Shifts via Phase-Amplitude Modulation for Non-stationary Time Series Forecasting (NeurIPS 2026 · similarity 0.87)
- TimeBase: The Power of Minimalism in Efficient Long-term Time Series Forecasting (ICML 2025 · similarity 0.87)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.87)
- OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025 · similarity 0.87)
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.
