Research map / Architectures and efficiency / RNNs and normalization
PulseImpute: A Novel Benchmark Task for Pulsative Physiological Signal Imputation
Maxwell Xu, Alexander Moreno, Supriya Nagesh, Varol Burak Aydemir, David W Wetter, Santosh Kumar, James Matthew Rehg
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 |
|---|---|---|---|
| Use BDC Transformer instead of the strongest reported baseline | ECG imputation and heartbeat detection · ECG imputation and heartbeat detection | MSE: .0276 ± .00019 -> .0194 ± .00017 | Improved |
| Use BDC Transformer instead of the strongest reported baseline | ECG imputation and heartbeat detection · ECG imputation and heartbeat detection | F1: .09 ± .001 -> .64 ± .003 | Improved |
| Use BDC Transformer instead of the strongest reported baseline | PPG imputation and heartbeat detection · PPG imputation and heartbeat detection | MSE: .0802 ± .00061 -> .0137 ± .00020 | Improved |
| Use BDC Transformer instead of the strongest reported baseline | PPG imputation and heartbeat detection · PPG imputation and heartbeat detection | F1: .25 ± .001 -> .81 ± .003 | Improved |
| Use BDC with RF=883 rather than vanilla self-attention with RF=1 | Transformer with parameters fixed at ~2.6 mil · BDC architecture comparison | MSE: 0.0177 -> 0.0123 | Improved |
| Use Conv Transformer with RF=9 instead of BDC with RF=883 (ablation) | Transformer with parameters fixed at ~2.6 mil · BDC architecture comparison | MSE: 0.0123 -> 0.0231 | Hurt |
| Add positional encoding (ablation) | BDC Transformer · Transformer positional encoding comparison | MSE: 0.0118 -> 0.0121 | Hurt |
| Add positional encoding (ablation) | Conv Transformer · Transformer positional encoding comparison | MSE: 0.0223 -> 0.0225 | Hurt |
| Use BDC instead of Conv Transformer at matched parameter count | Transformer with parameters fixed at ~2.6 mil · BDC architecture comparison | MSE: 0.0231 -> 0.0123 | Improved |
| Use BDC Transformer rather than FFT imputation | ECG imputation and heartbeat detection · ECG imputation and heartbeat detection | F1: .09 ± .001 -> .64 ± .003 | Improved |
Closest papers by meaning
- Training Recurrent Neural Networks with Inherent Missing Data for Wearable Device Applications (Student Abstract) (AAAI 2025 · similarity 0.86)
- Large-scale Training of Foundation Models for Wearable Biosignals (ICLR 2024 · similarity 0.85)
- Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions (ICML 2025 · similarity 0.85)
- OpenMHC: Accelerating the Science of Wearable Foundation Models (NeurIPS 2026 · similarity 0.84)
- PaPaGei: Open Foundation Models for Optical Physiological Signals (ICLR 2025 · similarity 0.83)
- PhysioWave: A Multi-Scale Wavelet-Transformer for Physiological Signal Representation (NeurIPS 2025 · 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.
