Research map / Time series / Foundation models for time series
Time Series as Images: Vision Transformer for Irregularly Sampled Time Series
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 ViTST (line graph images + Swin) instead of best specialized baseline | irregular time series classification · PAM | accuracy: 88.5 -> 95.8 | Helped |
| use ViTST vs Raindrop | irregular time series · P19 | AUROC: 87.0 -> 89.2 | Helped |
| ViTST vs TST on regular data | UEA datasets · UEA average accuracy | accuracy: 0.791 -> 0.780 | Hurt |
| remove variable-specific colors (ablation) | ViTST · P12 | AUPRC: 51.1 -> 47.0 | Hurt |
| remove markers on observed points (ablation) | ViTST · P19 | AUPRC: 53.1 -> 51.7 | Hurt |
| remove linear interpolation (plot observed points only) (ablation) | ViTST · P12 | AUROC: 85.1 -> 85.7 | Mixed |
| remove ordering variables by missing ratio (ablation) | ViTST · P12 | AUPRC: 51.1 -> 47.8 | Hurt |
| remove ImageNet pre-training (train Swin from scratch) (ablation) | ViTST · P19, P12, PAM | performance | Hurt |
| replace Swin/ViT with pre-trained ResNet (ablation) | ViTST · P19, P12, PAM | performance | Hurt |
| remove extreme values to narrow axis limits (IQR) (ablation) | ViTST · P19 | AUROC: 89.4 -> 88.2 | Hurt |
| vary line style/width and marker style/size (ablation) | ViTST · P19 | AUROC: 89.2 -> 88.5 to 89.3 | No clear effect |
| reduce overall image size at same grid layout (ablation) | ViTST · P12, PAM | performance | Hurt |
Closest papers by meaning
- A Diffusion Model for Regular Time Series Generation from Irregular Data with Completion and Masking (NeurIPS 2025 · similarity 0.89)
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 2025 · similarity 0.89)
- Utilizing Image Transforms and Diffusion Models for Generative Modeling of Short and Long Time Series (NeurIPS 2024 · similarity 0.88)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · similarity 0.85)
- Revitalizing Medical Time Series with Vision-Informed Retrieval: A Vision-Language Perspective (NeurIPS 2026 · similarity 0.85)
- Rough Transformers: Lightweight and Continuous Time Series Modelling through Signature Patching (NeurIPS 2024 · 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.