Research map / Time series / Long-term forecasting
Learning Latent Seasonal-Trend Representations for Time Series Forecasting
Zhiyuan Wang, Xovee Xu, Weifeng Zhang, Goce Trajcevski, Ting Zhong, Fan Zhou
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 seasonal component (ablation) | LaST · Synthetic, horizon 24 | MSE: 0.179 -> 0.129 | Improved |
| add trend component (ablation) | LaST · Synthetic, horizon 24 | MSE: 2.870 -> 0.129 | Improved |
| add trend component (ablation) | LaST · ETTh1 multivariate, horizon 24 | MSE: 0.631 -> 0.324 | Improved |
| add seasonal component (ablation) | LaST · ETTh1 multivariate, average | MSE: 0.631 -> 0.533 | Improved |
| add autocorrelation and CORT coefficients in reconstruction loss (ablation) | LaST · ETTh1 multivariate, horizon 24 | MSE: 0.334 -> 0.324 | Improved |
| add MI lower bound regularization (ablation) | LaST · ETTh1 multivariate, horizon 24 | MSE: 0.339 -> 0.324 | Improved |
| add MI upper bound regularization (ablation) | LaST · ETTh1 multivariate, horizon 24 | MSE: 0.332 -> 0.324 | Improved |
| use proposed unbiased MI lower bound instead of MINE (ablation) | LaST · ETTh1 multivariate, horizon 24 | MSE: 0.332 -> 0.324 | Improved |
| use proposed lower bound instead of MINE on ETTh1 horizon 720 (ablation) | LaST · ETTh1 multivariate, horizon 720 | MSE: 0.780 -> 0.740 | Improved |
| MINE vs no MI lower bound (w/o lower) on ETTh1 horizon 720 (ablation) | LaST · ETTh1 multivariate, horizon 720 | MSE: 0.766 -> 0.780 | Hurt |
| longer input length (96 to 672) for LaST at output 672 | LaST · ETTm1 multivariate | MSE: 0.532 -> 0.463 | Improved |
| longer input length (201 to 672) for LaST at output 48 | LaST · ETTm1 multivariate | MSE: 0.280 -> 0.279 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CoST — CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting (ICLR 2022) · better on 5 of 5 shared results
- Autoformer — Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting (NeurIPS 2021) · better on 5 of 5 shared results
- TS2Vec — TS2Vec: Towards Universal Representation of Time Series (AAAI 2022) · better on 2 of 2 shared results
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
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.88)
- OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025 · similarity 0.88)
- TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting (NeurIPS 2025 · similarity 0.87)
- Revitalizing Multivariate Time Series Forecasting: Learnable Decomposition with Inter-Series Dependencies and Intra-Series Variations Modeling (ICML 2024 · similarity 0.87)
- Self-Supervised Contrastive Learning for Long-term Forecasting (ICLR 2024 · similarity 0.87)
- TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024 · 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.
