Research map / Time series / Causal and generative time series
ContiFormer: Continuous-Time Transformer for Irregular Time Series Modeling
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 continuous-time attention with ODE-defined latent trajectories instead of discrete Transformer attention | ContiFormer · 2D spirals interpolation and extrapolation | RMSE: 1.37 -> 0.49 (interpolation); 1.36 -> 0.64 (extrapolation) | Helped |
| Use continuous-time attention with ODE-defined latent trajectories instead of discrete Transformer attention | ContiFormer · UEA classification, 30%, 50%, and 70% dropped | Avg. ACC.: 0.8089 -> 0.8126; 0.7793 -> 0.7997; 0.7401 -> 0.7749 | Helped |
| Use continuous-time attention for event prediction | ContiFormer · Traffic event prediction | LL: 0.372 ± .022 -> 0.635 ± .019$^+$ | Helped |
| Use continuous-time attention for event prediction | ContiFormer · MIMIC event prediction | LL: -1.110 ± .030 -> -1.135 ± .023 | Hurt |
| Use continuous-time Transformer modeling for regular forecasting | ContiFormer · ILI forecasting | MSE: 2.874 -> 2.632 | Helped |
| Use continuous-time Transformer modeling for regular forecasting | ContiFormer · Exchange forecasting | MSE: 0.277 -> 0.339 | Hurt |
| Use continuous-time Transformer modeling for regular forecasting | ContiFormer · Weather forecasting | MSE: 0.259 -> 0.257 | Helped |
| Use parallel continuous-time attention rather than recursive ODE-RNN integration | ContiFormer · RacketSports, 0% dropped | Time cost: 2.47 × -> 1.88 × (step size 0.1) | Helped |
Closest papers by meaning
- Rough Transformers: Lightweight and Continuous Time Series Modelling through Signature Patching (NeurIPS 2024 · similarity 0.87)
- Pathformer: Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting (ICLR 2024 · similarity 0.86)
- iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024 · similarity 0.85)
- DeformableTST: Transformer for Time Series Forecasting without Over-reliance on Patching (NeurIPS 2024 · similarity 0.85)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · similarity 0.85)
- Graph Neural Flows for Unveiling Systemic Interactions Among Irregularly Sampled Time Series (NeurIPS 2024 · similarity 0.85)
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.