- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 2
Read review
Paper summary
The author introduces a framework for learning representations from trajectories of SPD matrices called MatFAE. Instead of treating the data as discrete sequences, the authors model them as continuous functions. MatFAE combines geometry-aware layers, including multi-head BiMap, manifold pooling, and geodesic shrinkage, with a functional autoencoder to learn low-dimensional representations and reconstruct SPD trajectories. The authors also provide theoretical analysis of the proposed layers. The method is evaluated on ten simulation scenarios and six resting-state fMRI datasets, using clustering metrics such as AMI and ARI. MatFAE generally outperforms methods including SPDNet, TSRVF, GeoAtt, and PGA. The authors also test a supervised classification variant and conduct ablation experiments. However, the clustering scores on real fMRI datasets remain relatively low.
Critical strengths
- I think the combination of SPD geometry and functional data learning is an interesting direction. MatFAE provides a structured way to learn from SPD trajectories while preserving their geometric properties. The theoretical analysis of the proposed layers is also a useful part of the work.
- The experimental evaluation is fairly extensive. The authors test MatFAE on ten simulation scenarios and six real fMRI datasets, and compare it with several existing geometric learning methods. I particularly appreciate the simulation experiments, as they examine different aspects of the trajectories, such as smoothness, temporal ordering, and transition dynamics. The additional ablation and classification experiments also provide some evidence that the learned representations are useful beyond reconstruction.
Critical weaknesses
- My main concern is the relatively low clustering performance on the real fMRI datasets. Although MatFAE outperforms the baselines, the AMI scores are only between 0.030 and 0.137. This makes me wonder what the learned clusters actually represent. The paper discusses identifying meaningful patient subgroups, but I am not sure the current results are enough to support this. It would be interesting to see whether the clusters are related to clinical symptoms, cognitive measures, or other meaningful brain features, rather than just diagnostic labels.
- I also wonder how much the results depend on the sliding-window settings. The authors use 30-second windows with a 4-second step, which creates substantial overlap. Since temporal dynamics are central to this work, I think the authors should check whether different window sizes or steps lead to similar results.
Questions for authors
- The clustering results on real fMRI datasets are relatively low in terms of AMI and ARI. Beyond matching diagnostic labels, have the authors examined what the learned clusters actually represent? For example, are they associated with symptom severity, cognitive performance, or other clinical information?
- The dynamic FC matrices are estimated using 30-second sliding windows with a 4-second step. How sensitive are the results to these choices? I am curious whether longer windows or less overlap would change the clustering performance.
- Since the paper focuses on learning temporal information from SPD trajectories, have the authors compared MatFAE with a static FC baseline using one covariance matrix per subject? It would be useful to know how much the temporal information actually contributes beyond static functional connectivity.
- The paper reports improvements in AMI and ARI, but are these differences statistically significant across subjects or repeated runs? I also wonder how stable the clustering assignments are when using different random seeds or slightly different preprocessing settings.
6 comments
Response to Weakness 2 and Question 2: sensitivity analysis of the sliding window
We thank the reviewer for this suggestion. We agree that robustness to the sliding-window settings is important, because temporal dynamics are central to this work.
We note first that MatFAE treats each trajectory as a continuous function of time, so the windowed matrices are samples of a continuous trajectory. Overlapping windows are therefore the natural way to sample such a trajectory, and the stride mainly determines the sampling density.
To assess sensitivity, we repeated the COBRE analysis with window lengths from 30 s to 150 s (Table R11). Each window has 150 volumes, so longer windows also yield fewer windows per subject. We increased the stride for the longest windows accordingly.
Table R11. Sensitivity of MatFAE to window length and stride on COBRE (TR = 2 s). The number of clusters is fixed at their known value.
| Window (TR / s) | Stride (TR) | Windows per subject | MatFAE ARI |
|---|---|---|---|
| 15 / 30 (default) | 2 | 68 | 0.164 |
| 30 / 60 | 2 | 61 | 0.134 |
| 45 / 90 | 3 | 36 | 0.072 |
| 60 / 120 | 3 | 31 | 0.047 |
| 75 / 150 | 5 | 16 | 0.040 |
It can be seen that performance is stable across the commonly used range. Between 30 s and 60 s windows, ARI changes only from 0.137 to 0.134, so the results do not depend on the particular choice within this range. However, performance declines for long windows, as expected when temporal dynamics carry the signal. The drop in ARI shows that MatFAE exploits temporal structure rather than static connectivity. (As pointed out by Reviewer 2YVT, in our conclusion, we mistakenly compared with the AMI value (0.137), while the ARI value is 0.164. Table R11 shows that MatFAE performs best with the shortest tested window, and its ARI decreases consistently as window length increases.)
Choosing the window remains a modelling decision that depends on the data and the scientific question, for example the repetition time, the frequency band and the expected time scale of connectivity changes. Optimising it is beyond the scope of this work. Our Python package therefore takes the window length and stride as user-specified parameters.
Response to Question 3: comparison with static FC baseline
We thank the reviewer for this question. To isolate the contribution of temporal information, we compare full MatFAE with two types of controls:
-
Static FC baselines, which use one covariance matrix per subject computed over the entire run: tangent PCA followed by -means, and static SPDNet autoencoder.
-
Time-shuffled MatFAE, in which each subject's windows are randomly permuted before training. This keeps the set of visited connectivity states but destroys their order and rate of change.
Table R12. Mean ± standard deviation of ARI over 100 repetitions, with the number of clusters fixed at the true value. Full MatFAE values are from Table 3.
| Rung | Static PCA | Static SPDNet | MatFAE, time-shuffled | MatFAE (full) |
|---|---|---|---|---|
| A (temporal ordering) | -0.005 ± 0.008 | 0.000 ± 0.018 | -0.004 ± 0.005 | 1.000 ± 0.000 |
| B (static identity) | 0.231 ± 0.200 | -0.003 ± 0.009 | 0.425 ± 0.196 | 0.980 ± 0.009 |
| C (dwell time) | 0.384 ± 0.078 | 0.359 ± 0.067 | 0.432 ± 0.068 | 0.541 ± 0.032 |
| D (transition frequency) | -0.001 ± 0.008 | -0.004 ± 0.005 | -0.004 ± 0.007 | 0.286 ± 0.052 |
| E (smoothness) | 0.056 ± 0.044 | 0.157 ± 0.056 | 0.157 ± 0.077 | 0.808 ± 0.078 |
| F (Wishart concentration) | 0.056 ± 0.034 | 0.008 ± 0.022 | 0.999 ± 0.005 | 1.000 ± 0.000 |
| G (real HRF) | 0.040 ± 0.040 | -0.003 ± 0.009 | 0.042 ± 0.041 | 0.960 ± 0.018 |
| H (trajectory direction) | 0.004 ± 0.021 | 0.002 ± 0.015 | 0.008 ± 0.027 | 1.000 ± 0.000 |
| I (multi-scale dynamics) | 0.011 ± 0.036 | 0.011 ± 0.037 | 0.089 ± 0.064 | 0.085 ± 0.064 |
| J (Neurosynth) | 0.579 ± 0.258 | 0.632 ± 0.311 | 0.760 ± 0.168 | 1.000 ± 0.000 |
Table R13. ARI over 20 seeds
| Cohort | Static PCA | Static SPDNet | MatFAE, time-shuffled | MatFAE (full) |
|---|---|---|---|---|
| CNP | 0.005 ± 0.006 | 0.021 ± 0.007 | 0.002 ± 0.006 | 0.124 ± 0.019 |
| COBRE | 0.043 ± 0.005 | 0.063 ± 0.000 | 0.075 ± 0.013 | 0.164 ± 0.012 |
| ADHD-200 | 0.037 ± 0.013 | -0.007 ± 0.005 | 0.026 ± 0.024 | 0.103 ± 0.021 |
| ABIDE | -0.001 ± 0.001 | -0.002 ± 0.001 | -0.001 ± 0.001 | 0.039 ± 0.010 |
| TCP | 0.009 ± 0.002 | -0.010 ± 0.003 | 0.020 ± 0.013 | 0.055 ± 0.018 |
| CAT-D | -0.006 ± 0.000 | -0.007 ± 0.004 | 0.007 ± 0.013 | 0.079 ± 0.011 |
Full MatFAE outperforms both static baselines and the time-shuffled control on nine of the ten rungs. On rungs whose group differences are purely temporal (A, D, E, G, H), the static baselines and the time-shuffled control stay near chance (ARI ≤ 0.16). Where static connectivity carries part of the signal (C, J), temporal information still adds substantially (0.541 vs. ≤ 0.432 on C; 1.000 vs. ≤ 0.760 on J). On B and F, averaging over the whole run also removes non-temporal differences that time-resolved trajectories preserve: a briefly visited state on B, and connectivity concentration on F. The only exception is rung I, which remains difficult for all methods.
Response to Weakness 1 and Question 1: what the learned clusters represent
We thank the reviewer for this important question.
Data and setup. We performed this analysis on CNP, the cohort with the most comprehensive clinical characterisation: clinician-rated symptom scales, self-report questionnaires and a neuropsychological battery, across four diagnostic groups. We applied the same pipeline as detailed in the manuscript plus motion correction (global signal regression, exclusion of the 47 subjects with mean head motion above 0.25 mm (257 → 210), and motion regressed out of the connectivity features). MatFAE never had access to diagnosis, symptoms or cognition, so all associations below are post hoc.
Cluster composition. With stricter motion control, MatFAE's agreement with diagnosis is similar to that under the manuscript's pipeline (ARI 0.136 vs. 0.124 ± 0.019; Table R6).
Table R9. Diagnostic composition of the four MatFAE clusters (n=210). Enrichment odds ratios are from Fisher's exact test (Holm-corrected p<0.001).
The clusters are strongly associated with diagnosis (χ2(9)=125.1, p<0.001). The association is considerably stronger among patients alone (n=106; ARI 0.334, χ2(6)=90.5, p<0.001). C1, C3 and C4 each concentrate one patient group. The overall ARI is low mainly because controls do not form a cluster of their own: 97% of controls are spread over C1-C3. The low AMI and ARI therefore reflect how controls are distributed across clusters, not clusters unrelated to psychopathology.
Symptoms and cognition.
Table R10. Cluster means of clinical and cognitive composites, in SD units. Clinician-rated composites are standardised over the 106 patients; the other composites are standardised to controls (n=210). η2 is the variance explained; p is from the Kruskal-Wallis test. The last column is the F-test of adding clusters to diagnosis.
The clusters differ significantly in psychosis severity, self-reported symptoms and cognition, and all three differences remain significant after Holm correction across the four composites.
These differences largely follow diagnosis: adding the clusters to diagnosis does not significantly improve the fit for any composite (p≥0.115).
Connectivity beyond diagnosis. We compared the clusters on 28 static connectivity features: the mean Fisher-z connectivity within and between the seven Yeo networks (Schaefer-100 parcellation), with Benjamini-Hochberg FDR correction. Nine features differ between clusters (q<0.05), and five of these involve the default-mode network. C4 has the highest connectivity within the default-mode network (η2=0.109) and the lowest connectivity between the default-mode and dorsal-attention networks (η2=0.090), making its default-mode network the most segregated. Diagnosis explains at most 5% of the variance in any of these features, and adjusting for diagnosis changes little (partial η2 = 0.112 and 0.098; eight of the nine features remain at q<0.05). Because the clusters are learned from connectivity, these tests are descriptive rather than independent validation.
Summary. The clusters correspond to recognisable clinical profiles:
We will add this analysis to the appendix.