- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes PARSI for irregular multivariate time series forecasting. It uses parallel temporal branches to capture patterns from different historical ranges and an entropy-based selective attention mechanism to model cross-variable dependencies. Experiments on several IMTS benchmarks evaluate forecasting accuracy and computational efficiency.
Critical strengths
- The paper is generally clear and easy to follow.
- The experiments cover multiple IMTS datasets and recent baselines, with results reported over five runs.
Critical weaknesses
-
The novelty is somewhat limited. Parallel temporal modeling and adaptive fusion have been explored in previous multi-scale forecasting methods. The selective interaction module is also closely related to ProbSparse attention. The differences from recent IMTS methods such as APN, Hi-Patch, and ReIMTS could be explained more clearly.
-
The ablation study only tests what happens when the interaction module is removed. It does not compare entropy-based selection with random selection or full attention. These comparisons are needed to show whether the selection strategy actually helps.
-
In Supplementary Table 16, selective attention is slower than full attention for up to 1024 variables. Table 17 also shows that PARSI is much slower than t-PatchGNN at inference on several datasets. The efficiency advantage is therefore less clear for the datasets used in this paper.
-
Removing the cross-variable interaction module makes little difference on CESNET and ClusterTrace (Table 2). The changes are also small compared with the reported standard deviations, so the benefit of this module on these datasets is not convincing.
-
The temporal ablation compares different ranges, but does not compare parallel modeling with progressive multi-scale modeling. Such a comparison would help show whether using parallel branches is better than existing approaches.
Questions for authors
- Does entropy-based query selection outperform random selection or full attention under the same settings?
- Why is selective attention slower than full attention at the variable sizes used in most experiments?
- What specific advantage does parallel temporal modeling offer over existing multi-scale IMTS methods?
5 comments
Response to W3–W4, Q2
Response to W3 and Q2: Computational Efficiency of Selective Attention
We thank the reviewer for highlighting this issue. At smaller variable counts, full attention is already inexpensive, so the additional query-selection overhead outweighs the computational savings. However, as Table 3 shows, The advantage of selective attention emerges with large number of variables. Specifically, our controlled comparison shows that its computation time grows more slowly than full attention, yielding a 62.18% reduction in attention-core time at 4,096 variables. This is important for high-dimensional forecasting applications, such as large sensor networks and infrastructure monitoring.
The comparison with t-PatchGNN concerns the cost of the complete architecture, rather than the attention module alone. PARSI prioritizes forecasting accuracy through parallel temporal modeling and adaptive fusion, at the cost of higher inference latency. But our proposed method PARSI achieves lower MSE than t-PatchGNN on 8 of 9 evaluated datasets, and our ablations support the contribution of these architectural choices.
Thus, the longer runtime should not negate the demonstrated forecasting benefits. PARSI offers an accuracy–cost trade-off, with selective attention helping control cross-variable computation as the number of variables grows.
Table 2. Forecasting accuracy and inference latency comparison with t-PatchGNN
| Dataset | PARSI MSE ↓ | t-PatchGNN MSE ↓ | PARSI inference (ms/step) ↓ | t-PatchGNN inference (ms/step) ↓ |
|---|---|---|---|---|
| FNSPID | 0.0281 | 0.0388 | 7.81 | 1.71 |
| CESNET | 0.6742 | 0.6914 | 8.97 | 1.69 |
| ClusterTrace | 1.3655 | 1.8745 | 24.75 | 1.64 |
| USHCN | 0.1506 | 0.1772 | 39.48 | 4.02 |
Table 3. Scalability comparison of three cross-variable interaction strategies (ms/forward, mean ± SD)
| Variables | Selective cross-variable attention | Full cross-variable attention | Random state selection |
|---|---|---|---|
| 128 | 0.2811 ± 0.0043 | 0.1525 ± 0.0022 | 0.1710 ± 0.0007 |
| 256 | 0.2844 ± 0.0017 | 0.1448 ± 0.0030 | 0.1702 ± 0.0005 |
| 512 | 0.2801 ± 0.0051 | 0.1450 ± 0.0007 | 0.1739 ± 0.0051 |
| 1024 | 0.2828 ± 0.0014 | 0.1489 ± 0.0007 | 0.1734± 0.0007 |
| 2048 | 0.2910 ± 0.0015 | 0.2866 ± 0.0002 | 0.2084 ± 0.0023 |
| 4096 | 0.4112 ± 0.0013 | 1.0872 ± 0.0010 | 0.3674 ± 0.0005 |
Response to W4: Effectiveness of Cross-Variable Interaction
We thank the reviewer for this observation. Table 2 of the main paper evaluates the incremental contribution of each component within PARSI, rather than expecting every component to contribute equally on every dataset. The small gains on CESNET and ClusterTrace indicate a limited additional benefit from cross-variable interaction in these settings, not that the module is ineffective overall. For example, its removal has a larger effect on USHCN, particularly on MAE.
This dataset-dependent contribution is consistent with the relatively weak cross-variable correlations measured on CESNET and ClusterTrace. When cross-variable information provides limited additional predictive value, PARSI can derive most of its benefit from temporal modeling and adaptive fusion.
Thus, the ablations therefore clarify the complementary roles of PARSI’s components. Cross-variable interaction contributes differently across datasets, while the complete architecture achieves the lowest mean errors across the four datasets in Table 2 of the main paper. We will revise the discussion to distinguish this overall performance.
Response to W5 and Q3 : Advantages of Parallel Temporal Modeling
We thank the reviewer for suggesting this direct comparison. The main advantage of parallel temporal modeling is that each temporal branch independently accesses the original observations, preserving information across different temporal ranges before adaptive fusion. In contrast, progressive multi-scale methods such as Hi-Patch [2] and ReIMTS [5] construct representations sequentially, where later stages depend on previously aggregated information. This may compress useful temporal details before their relevance to forecasting is determined.
To isolate this design choice, we compare parallel and progressive temporal modeling within PARSI across 11 datasets and five paired seeds. As shown in Table 4 and 5, the results demonstrate the benefits of parallel temporal modeling under controlled conditions. Parallel encoding achieves lower MAE on 10 of 11 datasets and lower MSE on 7 of 11 datasets, while reducing inference latency on all 11 datasets.
Table 4. Forecasting Accuracy of Parallel and Progressive Temporal Modeling
| Dataset | Parallel MAE ↓ | Progressive MAE ↓ | Parallel MSE ↓ | Progressive MSE ↓ |
|---|---|---|---|---|
| FNSPID | 0.092053 ± 0.002515 | 0.092221 ± 0.002011 | 0.028007 ± 0.000407 | 0.028008 ± 0.000591 |
| StudentLife | 0.541387 ± 0.000933 | 0.542591 ± 0.001582 | 0.583221 ± 0.001267 | 0.584695 ± 0.001152 |
| EPA-Air | 0.481542 ± 0.005016 | 0.485880 ± 0.014627 | 0.413304 ± 0.009773 | 0.417551 ± 0.020137 |
| CESNET | 0.608555 ± 0.000750 | 0.608712 ± 0.001079 | 0.682584 ± 0.000631 | 0.682034 ± 0.001351 |
| ClusterTrace | 0.753991 ± 0.003625 | 0.757944 ± 0.003289 | 1.344334 ± 0.015757 | 1.346721 ± 0.011809 |
| RepoHealth | 0.674452 ± 0.029715 | 0.627628 ± 0.021348 | 1.520127 ± 0.084661 | 1.336648 ± 0.090141 |
| USHCN | 0.263026 ± 0.006087 | 0.264626 ± 0.010319 | 0.157706 ± 0.015350 | 0.168104 ± 0.013427 |
| P12 | 0.382088 ± 0.006814 | 0.523227 ± 0.086715 | 0.320642 ± 0.005168 | 0.512712 ± 0.147759 |
| HumanActivity | 0.122452 ± 0.000765 | 0.122731 ± 0.000754 | 0.045902 ± 0.000502 | 0.044680 ± 0.000772 |
| Baoding | 0.419466 ± 0.001432 | 0.419524 ± 0.001315 | 0.351095 ± 0.000470 | 0.350945 ± 0.000421 |
| Zhuzhou | 0.445357 ± 0.002019 | 0.445816 ± 0.002055 | 0.500849 ± 0.002184 | 0.501106 ± 0.002000 |
Table 5. Inference Latency of Parallel and Progressive Temporal Modeling (ms/batch, mean ± SD)
| Dataset | Parallel ↓ | Progressive ↓ |
|---|---|---|
| FNSPID | 7.121 ± 0.181 | 7.714 ± 0.259 |
| StudentLife | 8.229 ± 0.444 | 11.139 ± 1.159 |
| EPA-Air | 25.443 ± 2.395 | 26.687 ± 4.804 |
| CESNET | 9.492 ± 1.051 | 10.195 ± 0.920 |
| ClusterTrace | 23.629 ± 0.543 | 26.178 ± 3.866 |
| RepoHealth | 11.179 ± 0.178 | 12.194 ± 0.362 |
| USHCN | 48.016 ± 4.105 | 48.184 ± 4.732 |
| P12 | 8.605 ± 0.117 | 9.993 ± 1.338 |
| HumanActivity | 16.776 ± 2.140 | 17.411 ± 2.122 |
| Baoding | 13.373 ± 0.172 | 14.155 ± 0.110 |
| Zhuzhou | 12.086 ± 0.503 | 12.434 ± 0.102 |
Reference
[1] Wu, Y.-H., et al. STAnHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction. ICLR, 2024.
[2] Luo, Y., et al. Hi-Patch: Hierarchical Patch GNN for Irregular Multivariate Time Series. ICML, 2025.
[3] Zhou, H., et al. Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting. AAAI, 2021.
[4] Liu, X., et al. Rethinking Irregular Time Series Forecasting: A Simple Yet Effective Baseline. AAAI, 2026.
[5] Li, B., et al. Learning Recursive Multi-Scale Representations for Irregular Multivariate Time Series Forecasting. ICLR, 2026.
Response to W1–W2, Q1
Response to W1: Novelty of the Proposed Framework
We agree that multi-scale modeling, adaptive fusion, and sparse attention are established ideas. PARSI’s contribution is not these mechanisms individually, but how it preserves different historical ranges before compression and allocates cross-variable computation using the resulting representations. Furthermore, our contribution is their problem-driven adaptation and organization for IMTS forecasting. This architectural approach to methodological contribution is also illustrated by STanHop [1] and Hi-Patch [2].
Relationship to ProbSparse. PARSI shares ProbSparse’s principle of prioritizing queries with concentrated attention, and we will make this connection explicit [3]. ProbSparse selects temporal queries using an approximation based on the maximum and mean attention logits. PARSI uses sampled attention entropy to select variable states after multi-range temporal fusion. Selected states receive individualized cross-variable aggregation, while the others receive shared global context. The distinction is therefore in the selection criterion and its application to temporally summarized variable states, rather than a new general sparse-attention principle.
Differences from APN, Hi-Patch, and ReIMTS. APN [4] learns adaptive patch boundaries, while Hi-Patch and ReIMTS organize temporal representations hierarchically or recursively [5]. PARSI instead constructs independent temporal views directly from the same observation embeddings. Longer-range branches therefore do not depend on summaries already compressed by shorter-range branches. Feature-wise adaptive fusion then combines these complementary views into a representation for each variable. In our controlled comparison with progressive encoding, keeping adaptive fusion unchanged, parallel encoding reduces test MAE, test MSE, and inference latency by up to 26.97%, 37.46%, and 26.12%, respectively, across the evaluated settings.
Response to W2 and Q1: Effectiveness of Entropy-Based Query Selection
We thank the reviewer for this suggestion. We have added controlled comparisons with random selection and full attention across all 11 datasets, keeping the other PARSI components unchanged and using the same query budget for entropy-based and random selection.
When only a subset of queries is selected, entropy-based selection achieves lower mean MAE and MSE than random selection, supporting the benefit of informed selection. It also maintains comparable or better accuracy than full attention. On datasets where the budget covers all queries, the strategies reduce to full attention and produce identical results. These comparisons directly evaluate the selection strategy rather than only the presence of cross-variable interaction.
Note that entropy-based selection is one component of PARSI, not its sole contribution. As shown in Section 4.2 (Table 1) and runtime measurements in Section 4.4 (Figure 4), the careful design of PARSI’s architecture achieves better forecasting accuracy than several attention-based baselines while remaining computationally efficient.
Table 1. Forecasting performance of different cross-variable attention strategies (MAE/MSE, mean ± SD)