- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
This paper proposes SPAQR for irregular multivariate time-series forecasting, combining sampling-aware patch allocation with query-conditioned retrieval. The approach distributes a limited patch budget among variables based on their sampling patterns and represents each patch using coarse, boundary, and transition information. For each forecasting query, the model identifies and combines relevant historical information based on the target variable and prediction time. Experiments on four benchmark datasets against sixteen baseline methods show that SPAQR achieves the lowest mean MSE and MAE across all datasets. The ablation studies further support the effectiveness of its main components.
Critical strengths
1- The paper focuses on an important challenge in irregular time-series forecasting, where variables often have different sampling patterns and assigning the same number of patches to each variable may not be effective. Allocating a limited number of patches based on these sampling differences is an interesting and practical idea.
2- The idea of capturing different types of temporal information within each patch and selecting the relevant information based on the target variable and forecasting time is interesting. This allows the model to use the same historical data differently for different forecasting queries, rather than relying on a single fixed representation.
3- The approach is evaluated on four irregular time-series benchmarks against sixteen baseline methods, including several recent approaches. SPAQR achieves the lowest average forecasting errors across all four datasets. The ablation and sensitivity studies further demonstrate the effectiveness of the proposed framework and its main components.
Critical weaknesses
1- The proposed sampling-aware patch allocation is interesting, but the scoring function in Eq. (5) could be better justified. Although the authors use equal weights for observation count, temporal span, and sampling gaps, these factors have different numerical ranges and may not contribute equally to the final allocation score. I would suggest including an ablation study to examine the effect of each term individually and in combination. This would help explain which factors are most important for improving forecasting performance and whether all three are needed.
2- Query-conditioned retrieval is one of the main contributions of the paper. However, the proposed weighting mechanism in Eqs. (12)–(14) appears similar to standard attention, where relevant information is selected based on a query. It is therefore not clear what additional benefit the proposed retrieval mechanism provides over conventional cross-attention. Also, the ablation without query-conditioned retrieval still uses query-dependent weighting across patches, so it does not fully remove query conditioning. I would suggest comparing the proposed method with standard cross-attention using the same patch representations and experimental settings. This would help demonstrate the specific advantage of the proposed retrieval mechanism.
3- Can the authors provide a more detailed analysis of forecasting performance under different sampling densities and irregularity levels to directly demonstrate the advantages of sampling-aware allocation?
Questions for authors
Please address my all concerns raised in the weaknesses section.
5 comments
W2: Comparison with standard cross-attention and interpretation of the existing ablation.
1. The design concerns the compatibility score.
We agree that SPAQR's query-conditioned evidence retrieval is attention-based. Its score combines a coarse-support content term shared by all four states with state-specific content, temporal anchors, geometry, and target–source variable biases.
Using the notation of Eqs. (12)–(13), let denote the coarse-support content score plus . Then, over valid support–state pairs,
The factor in Eq. (13) cancels because every support has four states. This equivalence makes the relationship to standard attention explicit: our contribution is the construction of the support–state compatibility score, not a new normalization operation.
2. The existing ablation simplifies state reading.
The w/o query-conditioned retrieval does not remove all query conditioning. It uniformly averages the four states and removes the state-evidence term from the support score, while retaining query-dependent support weighting.
3. New cross-attention controls retain the four-state representation.
We replaced only the final joint support–state retrieval with standard dot-product cross-attention over all four states of every valid support. We also evaluated cross-attention + metadata, which adds the six-dimensional support geometry, source-variable identity, and state identity to the key features. The encoder architecture and training settings were unchanged, and each model was trained end to end.
| Final retrieval | HumanActivity MSE | HumanActivity MAE | USHCN MSE | USHCN MAE | PhysioNet’12 MSE | PhysioNet’12 MAE | MIMIC MSE | MIMIC MAE |
|---|---|---|---|---|---|---|---|---|
| Standard cross-attention | 0.0585 ± 0.0122 | 0.1587 ± 0.0244 | 0.1997 ± 0.0191 | 0.2834 ± 0.0046 | 0.3125 ± 0.0015 | 0.3738 ± 0.0040 | 0.4174 ± 0.0042 | 0.3863 ± 0.0055 |
| Cross-attention + metadata | 0.0506 ± 0.0041 | 0.1416 ± 0.0104 | 0.1869 ± 0.0053 | 0.2771 ± 0.0062 | 0.3124 ± 0.0017 | 0.3719 ± 0.0024 | 0.4274 ± 0.0048 | 0.3856 ± 0.0028 |
| SPAQR | 0.0409 ± 0.0007 | 0.1054 ± 0.0009 | 0.1582 ± 0.0009 | 0.2549 ± 0.0056 | 0.2899 ± 0.0016 | 0.3489 ± 0.0018 | 0.3956 ± 0.0018 | 0.3780 ± 0.0023 |
W3: Performance across sampling densities and irregularity levels.
1. Controlled changes to test-time sampling.
We compared sampling-aware patch allocation with Equal time from Appendix B.3, retaining the same encoding and retrieval architecture. Both models were trained on the original training data and evaluated without retraining. For each test sample, they received the same retained observations; forecast times and targets were unchanged. We also preserved the first and last observation of each nonempty variable.
2. Density, internal gaps, and variable imbalance.
- Density (50% nominal retention): Uniformly thin observations within each variable. Endpoint preservation and rounding may increase the realized retention fraction.
- Internal gaps (strength 1.0): Concentrate removals near a randomly chosen interior time, keeping the same per-variable retained counts as uniform thinning.
- Variable imbalance (strength 1.0): Vary retention quotas across variables while keeping the same total retained count per sample as uniform thinning.
3. Results under sampling perturbations.
The table reports MSE and MAE for the original history and three representative perturbations. Across the three reported perturbations, SPAQR achieves lower mean MSE on all four datasets and lower mean MAE in 10 of 12 comparisons.
| Condition | Allocation | HumanActivity MSE | HumanActivity MAE | USHCN MSE | USHCN MAE | PhysioNet’12 MSE | PhysioNet’12 MAE | MIMIC MSE | MIMIC MAE |
|---|---|---|---|---|---|---|---|---|---|
| Original history | SPAQR | 0.0409 ± 0.0007 | 0.1054 ± 0.0009 | 0.1582 ± 0.0009 | 0.2549 ± 0.0056 | 0.2899 ± 0.0016 | 0.3489 ± 0.0018 | 0.3956 ± 0.0018 | 0.3780 ± 0.0023 |
| Original history | Equal time | 0.0419 ± 0.0004 | 0.1179 ± 0.0008 | 0.1979 ± 0.0225 | 0.2745 ± 0.0077 | 0.3030 ± 0.0025 | 0.3640 ± 0.0049 | 0.4177 ± 0.0037 | 0.3861 ± 0.0049 |
| Uniform retention | SPAQR | 0.0493 ± 0.0013 | 0.1392 ± 0.0027 | 0.1814 ± 0.0139 | 0.2732 ± 0.0084 | 0.3076 ± 0.0033 | 0.3661 ± 0.0027 | 0.4130 ± 0.0046 | 0.3884 ± 0.0055 |
| Uniform retention | Equal time | 0.0508 ± 0.0040 | 0.1375 ± 0.0056 | 0.1964 ± 0.0156 | 0.2746 ± 0.0064 | 0.3086 ± 0.0015 | 0.3676 ± 0.0023 | 0.4142 ± 0.0042 | 0.3896 ± 0.0053 |
| Gap concentration | SPAQR | 0.0525 ± 0.0011 | 0.1436 ± 0.0035 | 0.1913 ± 0.0170 | 0.2785 ± 0.0135 | 0.3097 ± 0.0029 | 0.3660 ± 0.0024 | 0.4248 ± 0.0046 | 0.3913 ± 0.0053 |
| Gap concentration | Equal time | 0.0533 ± 0.0051 | 0.1436 ± 0.0074 | 0.2111 ± 0.0232 | 0.2862 ± 0.0113 | 0.3131 ± 0.0025 | 0.3712 ± 0.0032 | 0.4258 ± 0.0033 | 0.3925 ± 0.0065 |
| Variable imbalance | SPAQR | 0.0661 ± 0.0032 | 0.1635 ± 0.0034 | 0.1936 ± 0.0215 | 0.2853 ± 0.0144 | 0.3179 ± 0.0048 | 0.3709 ± 0.0032 | 0.4246 ± 0.0044 | 0.3890 ± 0.0051 |
| Variable imbalance | Equal time | 0.0693 ± 0.0213 | 0.1659 ± 0.0198 | 0.2167 ± 0.0266 | 0.2893 ± 0.0083 | 0.3211 ± 0.0029 | 0.3774 ± 0.0031 | 0.4287 ± 0.0036 | 0.3936 ± 0.0069 |
We thank the reviewer for recognizing the motivation for sampling-aware allocation and query-dependent use of historical evidence. Following your suggestions, we conducted three sets of additional experiments and clarify what each comparison establishes below.
W1: Rationale and contributions of the allocation-score terms.
1. Equal coefficients do not imply equal numerical contributions.
We agree. The coefficients in Eq. (5) specify a default weighting. The normalization and cap control the scales of the three terms, but do not equalize their numerical ranges or contributions.
2. Each term expresses a different split preference.
The three terms in Eq. (5) capture different aspects of sampling:
3. The full combination performs best in the score-term ablation.
We evaluated all seven nonempty combinations. The support budget, per-variable cap, multi-scale support encoding, query-conditioned evidence retrieval, and training settings were unchanged. The full combination achieves the lowest mean MSE and MAE on all four datasets.