Research map / Time series / Foundation models for time series
Repurposing Foundation Model for Generalizable Medical Time Series Classification
Nan Huang, Haishuai Wang, Zihuai He, Marinka Zitnik, Xiang Zhang
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 |
|---|---|---|---|
| replace task-specific CNN head with FORMED attention classifier (SDA, class embeddings, learnable queries) on TimesFM backbone (ablation) | TimesFM · ADFTD (3-class) | accuracy: 47.89 -> 85.19 | Improved |
| same change (ablation) | TimesFM · PTB-XL (5-class) | accuracy: 71.14 -> 82.73 | Improved |
| same change (ablation) | TimesFM · PTB (2-class) | accuracy: 92.17 -> 95.68 | Improved |
| same change (ablation) | TimesFM · TDBrain (2-class) | accuracy: 89.62 -> 88.94 | No clear effect |
| multi-dataset repurposing (FORMED) instead of per-dataset TSM training | FORMED · TDBrain (2-class) | accuracy: 89.62 -> 88.94 | No clear effect |
| use shared-backbone task-specific-head TSA models instead of per-dataset TSM training | TSA baselines · MedTS cohort | accuracy | Hurt |
| increase queries per class k at adaptation (ablation) | FORMED · ECG200 and StandWalkJump (unseen) | performance | Improved |
| use FORMED adapting only CEs and LQs vs TimesFM-TSA | FORMED · ECG200 and StandWalkJump (unseen) | performance | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Medformer — Medformer: A Multi-Granularity Patching Transformer for Medical Time-Series Classification (NeurIPS 2024) · better on 3 of 4 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective 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
- MIRA: Medical Time Series Foundation Model for Real-World Health Data (NeurIPS 2025 · similarity 0.87)
- Medformer: A Multi-Granularity Patching Transformer for Medical Time-Series Classification (NeurIPS 2024 · similarity 0.86)
- Revitalizing Medical Time Series with Vision-Informed Retrieval: A Vision-Language Perspective (NeurIPS 2026 · similarity 0.85)
- Knowledge-Empowered Dynamic Graph Network for Irregularly Sampled Medical Time Series (NeurIPS 2024 · similarity 0.83)
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 2025 · similarity 0.83)
- JoLT: Jointly Learned Representations of Language and Time-Series for Clinical Time-Series Interpretation (Student Abstract) (AAAI 2024 · similarity 0.83)
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.
