AI Paper Atlas

Research map / Time series / Causal and generative time series

What went wrong and when? Instance-wise feature importance for time-series black-box models

Sana Tonekaboni, Shalmali Joshi, Kieran Campbell, David K. Duvenaud, Anna Goldenberg

NeurIPS 2020 · Time series › Causal and generative time series · approach: Robust forecasting and evaluation · 26 citations (OpenAlex)

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 changedWhereMeasuredEffect
use FIT (temporal-shift-aware, generator-based counterfactuals) instead of gradient/attention/perturbation baselinesfeature importance for time series · State dataAUROC (explanation): 0.798 vs best baseline 0.565Improved
use FIT instead of baselinesfeature importance · Switch-feature dataAUROC (explanation): 0.720 vs AFO 0.590Improved
use FIT instead of baselines on a simple spike taskfeature importance · Spike dataAUPRC (explanation): 0.866 vs AFO 0.932Mixed
use FIT instead of baselines on MIMIC-mortalityfeature importance · MIMIC-mortality k=50AUROC drop: 0.133 vs FO 0.095Improved
use FIT instead of baselines on MIMIC-interventionfeature importance · MIMIC-interventionAUROC drop: 0.050 vs GradSHAP 0.072 (k=50); 0.024 vs 0.028 (95-pc)No clear effect
replace random-noise perturbations with generative-model counterfactuals (AFO to FIT) on Spikeperturbation-based importance · Spike dataAUROC (explanation): 0.942 -> 0.968Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy