Research map / Time series / Foundation models for time series
Self-Supervised Dynamical System Representations for Physiological Time-Series
Yenho Chen, Maxwell A Xu, James Matthew Rehg, Christopher John Rozell
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 time-varying parameters (ablation) | PULSE · Linear probe, average over HAR/ECG/PPG/EEG | accuracy: -7.62 when removed | Improved |
| use separate f_sys and f_init (ablation) | PULSE · Linear probe, average over HAR/ECG/PPG/EEG | accuracy: -1.22 when shared | Improved |
| add random crop for initial-condition inference (ablation) | PULSE · Linear probe, average over HAR/ECG/PPG/EEG | accuracy: -6.42 when removed | Improved |
| use pseudo-pairs instead of random pairs in cross-reconstruction (ablation) | PULSE · Linear probe, average over HAR/ECG/PPG/EEG | accuracy: -9.68 with random pairs | Improved |
| add time-varying parameters (ablation) | PULSE · Linear probe EEG | accuracy: -12.7 when removed | Improved |
| random pairs in cross-reconstruction (ablation) | PULSE · Linear probe ECG | accuracy: -22.99 | Hurt |
| train PULSE with positive-label oracle | PULSE · Synthetic sigma=5 | accuracy: 77.34 -> 82.65 | Improved |
| train PULSE with negative oracle | PULSE · Synthetic sigma=5 | accuracy: 77.34 -> 76.90 | Hurt |
| use PULSE instead of REBAR | linear probe · Linear probe HAR | accuracy: 95.35 -> 93.27 | Hurt |
| use PULSE instead of REBAR with 1% labels | semi-supervised · ECG 1% labels | accuracy: 67.56 -> 84.77 | Improved |
| use PULSE instead of REBAR | transfer · HAR to Gesture | accuracy: 78.17 -> 83.67 | Improved |
Closest papers by meaning
- Learning Interpretable Hierarchical Dynamical Systems Models from Time Series Data (ICLR 2025 · similarity 0.85)
- A scalable generative model for dynamical system reconstruction from neuroimaging data (NeurIPS 2024 · similarity 0.85)
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 2025 · similarity 0.84)
- Integrating Multimodal Data for Joint Generative Modeling of Complex Dynamics (ICML 2024 · similarity 0.84)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · similarity 0.84)
- Probabilistic Decomposed Linear Dynamical Systems for Robust Discovery of Latent Neural Dynamics (NeurIPS 2024 · similarity 0.84)
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.
