Research map / Time series / Causal and generative time series
Dynamic Time Warping Based Adversarial Framework for Time-Series Domain
Taha Belkhouja, Yan Yan, Janardhan Rao Doppa
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 |
|---|---|---|---|
| use DTW-AR adversarial training instead of standard training | 1D-CNN · Heartbeat | clean accuracy: 70% to 75% | Improved |
| use DTW-AR adversarial training instead of l2-based attack training | 1D-CNN · UCR multivariate | clean accuracy | Improved |
| use random alignment paths instead of the single optimal DTW path | DTW-AR · five UCR datasets | fraction of dissimilar adversarial examples: DTW threshold 0.01: 100% | Improved |
| set alpha_2 above zero (Euclidean-space penalty) (ablation) | DTW-AR attack · UCR multivariate | fooling rate | Improved |
| replace Soft-DTW (CW-SDTW) with random-path DTW-AR | DTW-AR · UCR multivariate | robustness / fooling rate | Improved |
| use random-path distance instead of exact DTW per iteration | DTW-AR · UCR multivariate | runtime per iteration | Improved |
| switch DTW variant from DTW_D to DTW_I (ablation) | DTW-AR · UCR multivariate | attack effectiveness | No clear effect |
| change norm in the DTW local distance among p=1,2,infinity (ablation) | DTW-AR · UCR multivariate | performance | No clear effect |
| train with resampling-warping augmentation instead of DTW-AR adversarial training | 1D-CNN · Epilepsy (DTW-AR adversarial examples) | accuracy: 96 -> 26 | Hurt |
| train with resampling-warping augmentation instead of DTW-AR adversarial training | 1D-CNN · Heartbeat (DTW-AR adversarial examples) | accuracy: 96 -> 20 | Hurt |
| use 1NN-DTW instead of 5NN or 10NN-DTW (ablation) | kNN-DTW · five UCR datasets | accuracy: Epilepsy 56 vs 40 and 32; Heartbeat 63 vs 67 and 68 | Mixed |
| use 1D-CNN instead of kNN-DTW on multivariate data | 1D-CNN · Epilepsy | accuracy: 56 -> 95 | Improved |
Closest papers by meaning
- Adversarial Attacks on Time Series (TPAMI 2020 · similarity 0.90)
- Fortifying Time Series: DTW-Certified Robust Anomaly Detection (NeurIPS 2025 · similarity 0.90)
- Training Robust Deep Models for Time-Series Domain: Novel Algorithms and Theoretical Analysis (AAAI 2022 · similarity 0.88)
- DTWNet: a Dynamic Time Warping Network (NeurIPS 2019 · similarity 0.87)
- Theoretical Analysis of Robust Overfitting for Wide DNNs: An NTK Approach (ICLR 2024 · similarity 0.84)
- ROME is Forged in Adversity: Robust Distilled Datasets via Information Bottleneck (ICML 2025 · similarity 0.83)
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.
