Research map / Time series / Causal and generative time series
Counterfactual eXplainable AI (XAI) Method for Deep Learning-Based Multivariate Time Series Classification
Alan Gabriel Paredes Cetina, Kaouther Benguessoum, Raoni Lourenco, Sylvain Kubler
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 the CAM-guided naive stage (ablation) | CONFETTI (FCN) · UEA MTS (7 datasets) | L1 distance: 145.22 -> 107.01 | Improved |
| add the CAM-guided naive stage (ablation) | CONFETTI (ResNet) · UEA MTS (7 datasets) | L1 distance: 168.13 -> 110.74 | Improved |
| add the CAM-guided naive stage (ablation) | CONFETTI (FCN) · UEA MTS (7 datasets) | execution time (s): 221.44 -> 162.76 | Improved |
| add the CAM-guided naive stage (ablation) | CONFETTI (ResNet) · UEA MTS (7 datasets) | execution time (s): 384.31 -> 302.84 | Improved |
| add the CAM-guided naive stage (ablation) | CONFETTI (FCN) · UEA MTS (7 datasets) | sparsity: 0.81 -> 0.84 | No clear effect |
| raise confidence threshold theta from 0.51 to 0.95 | CONFETTI alpha=0.5 (FCN) · UEA MTS (7 datasets) | counterfactual confidence: 0.69 -> 0.97 | Improved |
| raise confidence threshold theta from 0.51 to 0.95 | CONFETTI alpha=0.5 (FCN) · UEA MTS (7 datasets) | sparsity: 0.85 -> 0.81 | Hurt |
| lower alpha from 0.5 to 0.0 (favor sparsity) | CONFETTI theta=0.51 (FCN) · UEA MTS (7 datasets) | sparsity: 0.85 -> 0.88 | Improved |
| lower alpha from 0.5 to 0.0 (favor sparsity) | CONFETTI theta=0.51 (FCN) · UEA MTS (7 datasets) | counterfactual confidence: 0.69 -> 0.59 | Hurt |
| increase alpha from 0 to 1 | CONFETTI theta=0.51 · UEA MTS (7 datasets) | counterfactual confidence: 0.60 -> 0.80 | Improved |
| use CONFETTI instead of CoMTE | CoMTE (FCN) · UEA MTS (7 datasets) | sparsity: 0.54 -> 0.85 | Improved |
| use euclidean instead of DTW proximity distance in objective m3 (ablation) | CONFETTI · UEA MTS (7 datasets) | sparsity: 0.85 -> 0.84 | No clear effect |
Closest papers by meaning
- Counterfactual Explanations of Time Varying Rankings (Student Abstract) (AAAI 2025 · similarity 0.86)
- Robust Stochastic Graph Generator for Counterfactual Explanations (AAAI 2024 · similarity 0.85)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.84)
- DiCoFlex: Model-Agnostic Diverse Counterfactuals with Flexible Control (NeurIPS 2025 · similarity 0.84)
- Optimal Information Retention for Time-Series Explanations (ICML 2025 · similarity 0.84)
- F-Fidelity: A Robust Framework for Faithfulness Evaluation of Explainable AI (ICLR 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.
