Research map / Time series / Causal and generative time series
Multi-Time Attention Networks for Irregularly Sampled Time Series
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 learned time attention with mTAND-Full instead of L-ODE-ODE | PhysioNet interpolation · PhysioNet, 50%–90% observed | MSE (× 10^-3): 50%: 6.721 ± 0.109 -> 4.139 ± 0.029; 60%: 6.816 ± 0.045 -> 4.018 ± 0.048; 70%: 6.798 ± 0.143 -> 4.157 ± 0.053; 80%: 6.85 | Helped |
| use mTAND-Full instead of ODE-RNN | PhysioNet mortality classification · PhysioNet | AUC: 0.833 ± 0.009 -> 0.858 ± 0.004 | Helped |
| use the mTAND encoder without the full encoder-decoder framework (ablation) | mTAND-Full · PhysioNet mortality | AUC: 0.845 ± 0.004 -> 0.854 ± 0.001 | Helped |
| use mTAND-Full instead of ODE-RNN | MIMIC-III mortality classification · MIMIC-III | AUC: 0.8561 ± 0.0051 vs 0.8544 ± 0.0024 | No clear effect |
| use the mTAND encoder without the full encoder-decoder framework (ablation) | mTAND-Full · MIMIC-III mortality | AUC: 0.8544 ± 0.0024 -> 0.8419 ± 0.0017 | Hurt |
| use mTAND-Full instead of ODE-RNN | Human Activity classification · Human Activity | accuracy: 0.885 ± 0.008 -> 0.910 ± 0.002 | Helped |
| use the mTAND encoder without the full encoder-decoder framework (ablation) | mTAND-Full · Human Activity | accuracy: 0.910 ± 0.002 -> 0.907 ± 0.002 | Hurt |
| replace fixed positional encodings with learned continuous-time embeddings (ablation) | mTAND-Full · PhysioNet mortality | AUC: 0.845 ± 0.004 -> 0.858 ± 0.004 | Helped |
| replace fixed positional encodings with learned continuous-time embeddings (ablation) | mTAND-Full · MIMIC-III mortality | AUC: 0.843 ± 0.001 -> 0.854 ± 0.002 | Helped |
| replace the fixed RBF interpolation kernel with a learned mTAND similarity kernel (ablation) | IP-Nets-style classifier · PhysioNet mortality | AUC: 0.819 ± 0.006 -> 0.854 ± 0.001 | Helped |
| replace the fixed RBF interpolation kernel with a learned mTAND similarity kernel (ablation) | IP-Nets-style classifier · MIMIC-III mortality | AUC: 0.839 ± 0.001 -> 0.842 ± 0.001 | No clear effect |
| replace the fixed RBF interpolation kernel with a learned mTAND similarity kernel (ablation) | IP-Nets-style classifier · Human Activity | accuracy: 0.869 ± 0.007 -> 0.907 ± 0.002 | Helped |
Closest papers by meaning
- Interpolation-Prediction Networks for Irregularly Sampled Time Series (ICLR 2019 · similarity 0.90)
- MPTSNet: Integrating Multiscale Periodic Local Patterns and Global Dependencies for Multivariate Time Series Classification (AAAI 2025 · similarity 0.86)
- TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning (ICML 2024 · similarity 0.85)
- Revisiting Attention for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.84)
- Hi-Patch: Hierarchical Patch GNN for Irregular Multivariate Time Series (ICML 2025 · similarity 0.84)
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 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.