Research map / Time series / Long-term forecasting
T1: One-to-One Channel-Head Binding for Multivariate Time-Series Imputation
Dongik Park, Hyunwoo Ryu, Suahn Bae, Keondo Park, Hyung-Sin Kim
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 cross-variable attention instead of pointwise convolution (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.037 -> 0.033 | Improved |
| add cross-variable modeling at all (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.051 -> 0.033 | Improved |
| use one-to-one channel-head binding instead of 32 channels per head (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.037 -> 0.033 | Improved |
| use one-to-one binding instead of 16 channels per head (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.038 -> 0.033 | Improved |
| use one-to-one binding instead of 8 channels per head (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.035 -> 0.033 | Improved |
| add explicit mask channel to the input embedding (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.034 -> 0.033 | Improved |
| use PixelShuffle instead of linear upsampling (ablation) | T1 · ETT/Weather/Electricity avg | MSE: 0.034 -> 0.033 | Improved |
| use T1 instead of PatchTST at 70% missing | T1 · Point missing ratio 0.7 | MSE: 0.092 -> 0.049 | Improved |
| use T1 under block missing | T1 · Block missing | MSE: 0.050 -> 0.026 | Improved |
| use T1 on naturally missing data | T1 · PhysioNet2012 and AQI36 | MSE: 0.097 -> 0.075 (PhysioNet); 0.262 -> 0.226 (AQI36) | Improved |
| use T1 on Weather with MAE | T1 · Weather (point missing) | MSE: MAE 0.045 vs 0.042 (BRITS) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CSDI — CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation (NeurIPS 2021) · better on 2 of 2 shared results
- TimeMixer++ — TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024) · better on 1 of 1 shared result
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Frequency-aware Generative Models for Multivariate Time Series Imputation (NeurIPS 2024 · similarity 0.89)
- Discretizing Continuous Time Series for Imputation with Masked Diffusion Training (NeurIPS 2026 · similarity 0.89)
- ProCTI: Prototype-Refined Global Conditioning for Diffusion-Based Time Series Imputation (NeurIPS 2026 · similarity 0.89)
- Glocal Information Bottleneck for Time Series Imputation (NeurIPS 2025 · similarity 0.89)
- Diffusion Transformers for Imputation: Statistical Efficiency and Uncertainty Quantification (NeurIPS 2025 · similarity 0.88)
- Motif-aware Graph Neural Networks for Networked Time Series Imputation (AAAI 2025 · similarity 0.87)
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.
