Research map / Time series / Foundation models for time series
SimMTM: A Simple Pre-Training Framework for Masked Time-Series Modeling
Jiaxiang Dong, Haixu Wu, Haoran Zhang, Li Zhang, Jianmin Wang, Mingsheng Long
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 |
|---|---|---|---|
| SimMTM pre-training vs random initialization (ablation) | Transformer forecasting · Forecasting in-domain avg | MSE: 0.343 -> 0.308 | Improved |
| SimMTM vs Ti-MAE pre-training | forecasting · Forecasting in-domain avg | MSE: 0.336 -> 0.308 | Improved |
| SimMTM vs TF-C pre-training | classification · SleepEEG -> EMG | accuracy (%): 81.74 -> 97.56 | Improved |
| TST pre-training vs random init (masked baseline) | classification · Epilepsy -> Epilepsy | accuracy (%): 89.83 -> 80.89 | Hurt |
| add reconstruction loss and constraint loss (full SimMTM) vs random init (ablation) | SimMTM · SleepEEG -> EMG | accuracy (%): +9.7% and +16.0% absolute | Improved |
| add constraint loss (manifold structure in series-wise space) (ablation) | SimMTM · SleepEEG -> EMG | accuracy (%): +16.0% absolute | Improved |
| aggregate negatives too (PNSA) instead of positives only (PSA) (ablation) | SimMTM · ETTh1 in-domain | MSE: 0.420 -> 0.409 | Improved |
| aggregate negatives too (PNSA) instead of positives only (PSA) (ablation) | SimMTM · SleepEEG -> EMG | accuracy (%): 87.80 -> 97.56 | Improved |
| add SimMTM pre-training to PatchTST | PatchTST · ETTh1 | MSE: 0.417 -> 0.409 | Improved |
| add PatchTST sub-series masking pre-training | PatchTST · ETTh1 | MSE: 0.417 -> 0.430 | Hurt |
| use multiple masked series (M>1) instead of one (M=1) (ablation) | SimMTM · ETTh1 input-336-predict-96 | MSE | Improved |
| geometric masking instead of random masking (ablation) | SimMTM · ETTh1 | MSE | Improved |
Later papers that compared with this method
6 later accepted papers report SimMTM as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Dynamic Semantic Tokenization for Time Series via Elastic Sampling on Physics-aware Perception (AAAI 2026 · better on 5 of 5 shared results)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · better on 8 of 10 shared results)
- Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning (AAAI 2025 · better on 5 of 5 shared results)
- Content-aware Balanced Spectrum Encoding in Masked Modeling for Time Series Classification (AAAI 2025 · better on 1 of 4 shared results)
- Learning Without Augmenting: Unsupervised Time Series Representation Learning via Frame Projections (NeurIPS 2025 · better on 1 of 1 shared result)
- UP2ME: Univariate Pre-training to Multivariate Fine-tuning as a General-purpose Framework for Multivariate Time Series Analysis (ICML 2024 · better on 5 of 5 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CoST — CoST: Contrastive Learning of Disentangled Seasonal-Trend Representations for Time Series Forecasting (ICLR 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
- Language Pre-training Guided Masking Representation Learning for Time Series Classification (AAAI 2025 · similarity 0.88)
- Learning to Embed Time Series Patches Independently (ICLR 2024 · similarity 0.87)
- TimeSiam: A Pre-Training Framework for Siamese Time-Series Modeling (ICML 2024 · similarity 0.87)
- Content-aware Balanced Spectrum Encoding in Masked Modeling for Time Series Classification (AAAI 2025 · similarity 0.87)
- Enhancing Masked Time-Series Modeling via Dropping Patches (AAAI 2025 · similarity 0.86)
- Discretizing Continuous Time Series for Imputation with Masked Diffusion Training (NeurIPS 2026 · similarity 0.86)
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.
