Research map / Time series / Causal and generative time series
Temporal Dependencies in Feature Importance for Time Series Prediction
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 WinIT instead of the strongest listed baseline, AFO | feature importance on Spike · Spike | AUPRC: 0.944_{\pm0.01} -> 1.0_{\pm0.0} | Helped |
| use WinIT instead of Dynamask | feature importance on Delayed-Spike · Delayed-Spike | AUPRC: 0.502_{\pm0.002} -> 0.996_{\pm0.009} | Helped |
| use WinIT instead of Dynamask | feature importance on State · State | AUPRC: \textbf{0.278}_{\pm0.003} -> 0.26_{\pm0.01} | Hurt |
| use WinIT instead of Dynamask | feature importance on MIMIC-III mortality · MIMIC-III mortality, top 5% | AUC Drop: 0.068_{\pm0.004} -> \textbf{0.098}_{\pm0.008} | Helped |
| use WinIT instead of Dynamask | feature importance on MIMIC-III mortality · MIMIC-III mortality, k=50 | Pred. Change: 0.076_{\pm0.004} -> \textbf{0.079}_{\pm0.004} | Helped |
| use KL divergence instead of prediction difference (ablation) | WinIT · MIMIC-III mortality, STD masking, top 5% | AUC Drop: 0.078_{\pm0.008} -> 0.076_{\pm0.006} | Hurt |
| use KL divergence instead of prediction difference (ablation) | WinIT · MIMIC-III mortality, STD masking, k=50 | AUC Drop: 0.076_{\pm0.006} -> 0.072_{\pm0.007} | Hurt |
| use Jensen-Shannon divergence instead of prediction difference (ablation) | WinIT · MIMIC-III mortality, STD masking, top 5% | AUC Drop: 0.078_{\pm0.008} -> 0.075_{\pm0.007} | Hurt |
| use Jensen-Shannon divergence instead of prediction difference (ablation) | WinIT · MIMIC-III mortality, STD masking, k=50 | AUC Drop: 0.076_{\pm0.006} -> 0.072_{\pm0.007} | Hurt |
| replace mean aggregation with max aggregation (ablation) | WinIT · MIMIC-III mortality, STD masking, top 5% | AUC Drop: 0.078_{\pm0.008} -> 0.066_{\pm0.008} | Hurt |
| replace mean aggregation with max aggregation (ablation) | WinIT · MIMIC-III mortality, STD masking, k=50 | AUC Drop: 0.076_{\pm0.006} -> 0.063_{\pm0.004} | Hurt |
| increase the WinIT window size from 1 to 3 (ablation) | WinIT · Delayed-Spike | AUPRC | Helped |
Closest papers by meaning
- ShapeX: Shapelet-Driven Post Hoc Explanations for Time Series Classification Models (NeurIPS 2025 · similarity 0.85)
- Explaining Time Series via Contrastive and Locally Sparse Perturbations (ICLR 2024 · similarity 0.85)
- Explain Temporal Black-Box Models via Functional Decomposition (ICML 2024 · similarity 0.84)
- Optimal Information Retention for Time-Series Explanations (ICML 2025 · similarity 0.84)
- TIMING: Temporality-Aware Integrated Gradients for Time Series Explanation (ICML 2025 · similarity 0.84)
- InteDisUX: Intepretation-Guided Discriminative User-Centric Explanation for Time Series (AAAI 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.