Research map / Time series / Foundation models for time series
BioFormer: Rethinking Cross-Subject Generalization via Spectral Structural Alignment in Biomedical Time-Series
Guikang Du, Haoran Li, Xinyu Liu, Zhibo Zhang, Xiaoli Gong, Jin 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 |
|---|---|---|---|
| add FBAM (spectral alignment) and SCLN (ablation) | Transformer backbone · APAVA | F1 (%): 74.23 -> 79.77 | Improved |
| add FBAM (spectral alignment) (ablation) | BioFormer · ADFTD | F1 (%): 45.90 -> 53.82 | Improved |
| add SCLN on top of FBAM (ablation) | BioFormer · APAVA | F1 (%): 78.07 -> 79.77 | Improved |
| add SCLN on top of FBAM (ablation) | BioFormer · PTB-XL | F1 (%): 62.56 vs 62.53 | No clear effect |
| use phase-only modulation instead of magnitude+phase (ablation) | BioFormer · PTB | F1 (%): 83.06 -> 84.71 | Improved |
| use magnitude-only modulation instead of joint (ablation) | BioFormer · APAVA | F1 (%): 79.77 -> 75.82 | Hurt |
| align in frequency domain (FBAM) instead of time domain (TSAM) (ablation) | Transformer · APAVA | F1 (%): 76.71 -> 80.00 | Improved |
| align in frequency domain (FBAM) instead of time domain (TSAM) (ablation) | Transformer · ADFTD | F1 (%): 47.45 -> 52.13 | Improved |
| align in frequency domain (FBAM) instead of time domain (TSAM) (ablation) | Transformer · PTB-XL | F1 (%): 58.86 -> 60.90 | Improved |
| use FBAM instead of GRL/MMD/CORAL/SubjectNorm/MixStyle/DSU (ablation) | Medformer · APAVA | F1 (%): 76.21 -> 80.00 (Trans. GRL vs FBAM column) | Improved |
| add SubjectNorm (ablation) | Transformer · APAVA | accuracy (%): 73.12 -> 47.11 | Hurt |
| add FBAM to Medformer (ablation) | Medformer · APAVA | F1 (%): 76.02 -> 81.46 | Improved |
Closest papers by meaning
- BIOT: Biosignal Transformer for Cross-data Learning in the Wild (NeurIPS 2023 · similarity 0.85)
- Medformer: A Multi-Granularity Patching Transformer for Medical Time-Series Classification (NeurIPS 2024 · similarity 0.83)
- Learning Robust Spectral Dynamics for Temporal Domain Generalization (NeurIPS 2025 · similarity 0.83)
- Spectral Transformer Neural Processes (NeurIPS 2026 · similarity 0.83)
- A scalable generative model for dynamical system reconstruction from neuroimaging data (NeurIPS 2024 · similarity 0.82)
- Population Transformer: Learning Population-level Representations of Neural Activity (ICLR 2025 · similarity 0.82)
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.
