Research map / Time series / Long-term forecasting
QuITE: Query-Based Irregular Time Series Embedding
Junghoon Lim
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 QuITE query-based embedding | iTransformer · IMTS forecasting (average) | MSE: 8.37 -> 3.79 | Improved |
| add QuITE query-based embedding | PatchTST · IMTS forecasting (average) | MSE: 7.59 -> 6.36 | Improved |
| add QuITE query-based embedding | PatchMixer · IMTS forecasting (average) | MSE: 6.84 -> 6.49 | Improved |
| add QuITE query-based embedding | PatchMixer · IMTS classification (average) | average score: 65.2 -> 75.5 | Improved |
| add QuITE query-based embedding | TMix · IMTS classification P19 AUPRC | AUPRC: 10.1 -> 38.0 | Improved |
| add QuITE to variate-token models rather than patch-token models | backbones · IMTS forecasting | MSE | Mixed |
| use query-token aggregation instead of mean pooling after self-attention (ablation) | iTransformer · PhysioNet horizon-average | MSE: 12.15 -> 5.21 | Improved |
| use query-token aggregation instead of mTAND grid embedding (ablation) | iTransformer · PhysioNet horizon-average | MSE: 13.11 -> 5.21 | Improved |
| add time embedding to values (Add) or concatenate timestamps (Concat) instead of QuITE (ablation) | iTransformer · PhysioNet horizon-average | MSE: 18.26 / 18.27 vs 5.21 | Hurt |
| vary query token initialization (Xavier, uniform, zero, random) (ablation) | QuITE++ · Human Activity / USHCN | MSE: 2.45-2.46 MSE on Human Activity | No clear effect |
| remove 75% additional observations at test time | QuITE++ · Human Activity, USHCN, PhysioNet | MSE | Hurt |
| remove up to 50% additional observations | QuITE++ · Human Activity, USHCN, PhysioNet | MSE | No clear effect |
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.
- HyperIMTS — HyperIMTS: Hypergraph Neural Network for Irregular Multivariate Time Series Forecasting (ICML 2025) · better on 2 of 2 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 2 of 2 shared results
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 2 of 2 shared results
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
- TimeCHEAT: A Channel Harmony Strategy for Irregularly Sampled Multivariate Time Series Analysis (AAAI 2025 · similarity 0.89)
- TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment (AAAI 2025 · similarity 0.88)
- HDT: Hierarchical Discrete Transformer for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.88)
- HyperIMTS: Hypergraph Neural Network for Irregular Multivariate Time Series Forecasting (ICML 2025 · similarity 0.87)
- Temporal Query Network for Efficient Multivariate Time Series Forecasting (ICML 2025 · similarity 0.87)
- Irregular Multivariate Time Series Forecasting: A Transformable Patching Graph Neural Networks Approach (ICML 2024 · 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.
