AI Paper Atlas

Research map / Time series / Causal and generative time series

Why Did This Model Forecast This Future? Information-Theoretic Saliency for Counterfactual Explanations of Probabilistic Regression Models

Chirag Raman, Alec Nonnemaker, Amelia Villegas-Morcillo, Hayley Hung, Marco Loog

NeurIPS 2023 · Time series › Causal and generative time series · approach: Probabilistic and conformal forecasting

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
replace DeepSHAP with the entropy-based saliency frameworksaliency computation for forecasts · Group leaving (MatchNMingle)execution time: 13.5 min -> 2 msImproved
replace DeepSHAP with the entropy-based saliency frameworksaliency computation for forecasts · Vehicle trajectory (nuScenes)execution time: 26.3 min -> 33 msImproved
use perfect forecasting model with entropy saliencysaliency framework · synthetic turn-taking dataidentifies true salient timesteps 138 and 139Improved
use real Social Process forecaster instead of perfect forecastssaliency framework · synthetic turn-taking dataidentifies true salient timestepsImproved
compare against DeepSHAP, GradientSHAP, IntegratedGradients, SmoothGradattribution on forecast mean and std · synthetic turn-taking datacapturing salient predictive relationshipsHurt
compare against DeepSHAP and GradientSHAPattribution on forecast mean and std · MatchNMingle group leavingintuitive patternsNo clear effect

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