Research map / Time series / Foundation models for time series
Unified Time Series Explanations via Amortized Optimization and Instance-level Multi-Expert Knowledge Distillation
Viet-Hung Tran, Zichi Zhang, Ngoc Phu Doan, Xuan Hoang Nguyen, Phi Hung Nguyen, Yimeng An, Peixin Li, Hans Vandierendonck, Ira Assent, Son Thai Mai
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 expert distillation to metric-loss training (ablation) | XMA explainer · MIT-ECG | AUPRC: 0.429 -> 0.890 | Improved |
| add expert distillation to metric-loss training (ablation) | XMA explainer · MCCE | AUPRC: 0.322 -> 0.889 | Improved |
| add metric loss to distillation-only training (ablation) | XMA explainer · MIT-ECG | AUPRC: 0.875 -> 0.890 | Improved |
| add metric loss to distillation-only training (ablation) | XMA explainer · MCCE | AUPRC: 0.773 -> 0.889 | Improved |
| add metric loss to distillation-only training (ablation) | XMA explainer · UEA (11 datasets) | faithfulness: higher on 8 of 11 datasets | Improved |
| add metric loss to distillation-only training (ablation) | XMA explainer · UEA (11 datasets) | robustness: higher across all tasks | Improved |
| add gradient (Input x Gradient) input to the explainer (ablation) | XMA explainer · UEA (11 datasets) | faithfulness | Improved |
| add gradient (Input x Gradient) input to the explainer (ablation) | XMA explainer · UEA (11 datasets) | robustness: Input-Only higher robustness on 10 of 11 datasets | Hurt |
| replace per-instance optimization (MIX) with amortized explainer | MIT-ECG explanations · MIT-ECG (200 instances) | explanation time (s): 152.3 -> 0.23 | Improved |
| use XMA instead of MIX | MIT-ECG explanations · MIT-ECG | AUPRC: improving AUPRC by ~0.14 | Improved |
Closest papers by meaning
- Optimal Information Retention for Time-Series Explanations (ICML 2025 · similarity 0.87)
- InteDisUX: Intepretation-Guided Discriminative User-Centric Explanation for Time Series (AAAI 2025 · similarity 0.86)
- Inherently Interpretable Time Series Classification via Multiple Instance Learning (ICLR 2024 · similarity 0.86)
- TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop (NeurIPS 2025 · similarity 0.86)
- TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning (ICML 2024 · similarity 0.85)
- TIMING: Temporality-Aware Integrated Gradients for Time Series Explanation (ICML 2025 · 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.
