Research map / Time series / Foundation models for time series
Learning a Warping Distance from Unlabeled Time Series Using Sequence Autoencoders
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 Autowarp’s learned distance instead of SSPD | Taxicab Mobility · Taxicab Mobility | normalized distance to nearest neighbors | Helped |
| Use Autowarp’s learned distance instead of EDR | Australian Sign Language (ASL) · ASL with added noise | nearest-neighbor classification accuracy | Helped |
| Minimize latent betaCV when selecting a warping distance | Libras · Libras | nearest-neighbor label accuracy: rank correlation is =0.85 | Helped |
| Use the sequence autoencoder’s latent representations alone, without warping-distance search (ablation) | Autowarp · Synthetic trajectories | clustering quality | Hurt |
| Search for a warping distance in the original trajectory space, without the autoencoder’s latent representations (ablation) | Autowarp · Synthetic trajectories | metric-learning quality | Hurt |
| Increase the sequence autoencoder to three layers (ablation) | Autowarp · Synthetic trajectories | similarity structure in latent representations | No clear effect |
| Vary the latent dimensionality (ablation) | Autowarp · Trajectories with outliers | qualitative ranking of warping distances | No clear effect |
| Vary the latent betaCV threshold percentile (ablation) | Autowarp · Trajectories with outliers | qualitative ranking of warping distances | No clear effect |
| Train the autoencoder for more batches (ablation) | Autowarp · Synthetic trajectories | latent betaCV | Helped |
Closest papers by meaning
- Perturbed Dynamic Time Warping: A Probabilistic Framework and Generalized Variants (ICLR 2026 · similarity 0.88)
- Deep Canonical Time Warping (CVPR 2016 · similarity 0.87)
- TimePoint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning (ICML 2025 · similarity 0.85)
- Improved Learning via k-DTW: A Novel Dissimilarity Measure for Curves (ICML 2025 · similarity 0.83)
- Learning Without Augmenting: Unsupervised Time Series Representation Learning via Frame Projections (NeurIPS 2025 · similarity 0.82)
- STEP: Learning STructured Embeddings for Progressive Time Series (NeurIPS 2026 · similarity 0.82)
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.