- Rating
- 3
- Confidence
- 3
- Informativeness
- 3
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes QMLR, a new classifier for pseudo-hyperbolic manifolds with multiple timelike dimensions. It combines spacelike and timelike decision scores using a learnable gate. This study also introduces a pseudo-hyperbolic attention module to improve feature representations while keeping them on the manifold.
Critical strengths
- This study roposes a new classifier based on pseudo-hyperbolic geometry, which is an interesting and less explored direction in geometric deep learning.
- The proposed method is evaluated on several different tasks, including graph classification, EEG decoding, image classification, and genomic sequence learning, showing its potential for use in different application areas.
- This paper also includes ablation studies and comparisons with existing methods.
Critical weaknesses
- The timelike branch uses a surrogate score instead of a well-defined minimum-distance formulation, which makes its geometric meaning less clear.
- The performance improvements are not consistent across all datasets, and QMLR does not always outperform existing methods.
- Some design choices, such as the number of timelike and spatial dimensions, may strongly affect performance, but their sensitivity and general impact need further investigation.
Questions for authors
- How sensitive is QMLR to the choice of timelike and spatial dimensions, and how should these values be selected in practice?
- Can the author/authors provide a stronger theoretical explanation for why the proposed timelike surrogate score is suitable for classification?
- Why does QMLR perform worse than some baseline methods on certain genomic sequence tasks?
- Can you provide more analysis to explain why the proposed method gives clear improvements on some datasets but only small or no improvements on others?
- How well does QMLR scale to larger datasets and deeper models, and what additional computational cost does it introduce compared with standard multinomial logistic regression?
3 comments
Response to Reviewer uZdY - Part 2
Q3–4. Dataset-dependent improvements
-
Dataset dependence. We agree that QMLR’s benefits vary across datasets. In these experiments, LMLR, BMLR-L, and QMLR share the CNN architecture and experimental settings, with only the classification head changed. The results show that pseudo-hyperboloid and QMLR is useful on most tasks. For the EEG tasks, we further examine input conditioning and geometric structure to explore possible reasons for the performance differences in that domain.
-
Comparison with matrix-manifold methods. We have examined input correlation conditioning to investigate the performance differences between Q-Net and CorAtt. Experiment shows that ERN has the largest condition number among the three datasets, suggesting greater sensitivity to matrix operations involving small eigenvalues. Q-Net learns pseudo-hyperbolic embeddings from encoded EEG features, providing an alternative that may be useful in this setting.
Dataset Correlation condition number Q-Net Vs CorAtt (%) MI 0.70↓ SSVEP 0.36↓ ERN 3.58↑ Larger condition numbers indicate poorer numerical conditioning. Performance differences use accuracy for MI/SSVEP and ROC-AUC for ERN.
-
Comparison with hyperbolic models. We also use sampled four-point hyperbolicity to examine waveform geometry. The analysis shows that ERN has the largest relative hyperbolicity value in waveform space, indicating less tree-like metric structure under this measure. Pseudo-hyperbolic geometry provides angular directions alongside hyperbolic slices, offering additional flexibility when relationships extend beyond hierarchy.
MI SSVEP ERN Here , where is estimated from sampled four-point maxima and is the sampled point cloud’s diameter. The reported dispersion is the between-subject standard deviation after averaging sampling repeats. This observation motivates considering representations beyond purely hierarchical geometry. Pseudo-hyperbolic space offers an option by combining hyperbolic slices with angular directions.
-
Manifold selection. These comparisons illustrate why no single geometry should be expected to perform best across all datasets, the following table gives practical starting points by our suggestion
Data structure Geometry Motivation Hierarchical Hyperbolic Captures hierarchy. Periodic or cyclic Circle / sphere Encodes angular variation. Mixed hierarchy and cycles Pseudo-hyperbolic Combines hyperbolic and angular structure. Informative covariance SPD Preserves variances and dependencies. Correlation with nuisance scaling Correlation manifold Removes positive scaling effects. Subspace-based Grassmann Represents basis-invariant subspaces.
Response to Reviewer uZdY - Part 3
Q5. Scalability and computational cost
-
Asymptotic cost: For classes and representation dimension , QMLR evaluates class projections and scalar score transformations in time per example, the same asymptotic order as Euclidean MLR. Norms, hyperbolic functions, and elementwise operations add a constant computational overhead.
-
Training cost: Table 5 reports training time and classifier parameter counts under the shared ResNet-18 setup.
Method CIFAR-10 Time CIFAR-10 Params CIFAR-100 Time CIFAR-100 Params MLR 14.2 5.13K 14.1 51.30K PMLR 16.3 5.14K 33.8 51.40K PBMLR 17.1 10.24K 37.9 102.40K BMLR-P 14.5 5.14K 14.5 51.40K LMLR 14.5 5.13K 14.5 51.30K BMLR-L 14.4 5.14K 14.3 51.40K QMLR 15.0 5.15K 14.9 51.50K Time: seconds per epoch. Params: classifier parameter count.
We additionally evaluate training cost in node classification datasets.
Dataset Readout Params Inference (ms) Fwd+bwd (ms) Inference peak memory (MiB) Fwd+bwd peak memory (MiB) Cora log+Linear 133 4.074 9.871 10.594 19.460 Cora QMLR 134 0.798 2.901 8.996 17.416 Airport log+Linear 72 4.195 9.109 10.909 19.829 Airport QMLR 73 0.810 2.882 8.829 17.220 CiteSeer log+Linear 108 4.146 9.207 10.999 19.978 CiteSeer QMLR 109 0.784 2.895 9.060 17.494 Arxiv log+Linear 720 6.682 20.814 158.540 233.576 Arxiv QMLR 721 3.964 14.456 255.104 367.233 -
Pretrained backbones: To evaluate Q-Net on larger datasets and with deeper networks, we replace its Euclidean feature extractor with pretrained LaBraM or EEGPT and evaluate it on BCIC2a, TUAB, and TUEV. The encoder’s output tokens are mapped to the pseudo-hyperbolic manifold via the exponential map, followed by Q-Net’s attention, residual refinement, and QMLR classifier. We fine-tune the encoder jointly with these downstream components. Results are summarized in Section 5.2 and Table 4, with experimental details and additional metrics provided in Appendix D.1.7.
BCIC2a
Method Balanced accuracy F1 Cohen's LaBraM LaBraM + Q-Net EEGPT EEGPT + Q-Net TUAB
Method Balanced accuracy AUROC PRAUC LaBraM LaBraM + Q-Net EEGPT EEGPT + Q-Net TUEV
Method Balanced accuracy AUROC PRAUC LaBraM LaBraM + Q-Net EEGPT EEGPT + Q-Net
References
- [a] Xiong, B., et al. (2022). Pseudo-Riemannian Graph Convolutional Networks. NeurIPS.
- [b] Le, V. Q., and Ta, V. C. (2025). Pseudo-Riemannian Graph Transformer. NeurIPS.
- [c] Jiang, W.-B., Zhao, L., and Lu, B.-L. (2024). Large Brain Model for Learning Generic Representations with Tremendous EEG Data in BCI. ICLR.
- [d] Wang, G., et al. (2024). EEGPT: Pretrained Transformer for Universal and Reliable Representation of EEG Signals. NeurIPS.
- [e] Xiong, W., et al. (2025). EEG-FM-Bench: A Comprehensive Benchmark for the Systematic Evaluation of EEG Foundation Models. arXiv:2508.17742.
Response to Reviewer uZdY - Part 1
We thank Reviewer uZdy for the valuable comments.
Q1. Sensitivity to timelike and spatial dimensions
The signature affects performance substantially, and the preferred allocation differs between datasets. We therefore treat (t,s) as a validation-selected hyperparameter rather than prescribe a single signature.
Geometric motivation: At fixed total dimension d=t+s, changing (t,s) changes the balance between temporal angular directions and spatial hyperbolic directions. The constructions in [a,b] motivate considering more angular temporal directions for cyclic relations and more spatial directions for hierarchical relations.
Sensitivity comparison: Figure 3 uses dataset-specific signatures motivated by [a] and keeps them fixed across spacelike, timelike, and fused heads. The sensitivity analysis is shown below.
The reported results favor different signatures on the two datasets: (14,2) on Cora and (1,15) on Airport. Performance is not totally monotonic in the number of timelike directions, so validation-based selection is needed. The procedure in [b, Section 5] compares estimated graph sectional-curvature distributions with reference distributions and offers a geometry-based starting point. Appendix D.3.3 shows the sensitivity analysis and selection guidelines, referenced in Section 5.1.
Q2. Suitability of the timelike surrogate
- Timelike distance. We appreciate the reviewer’s request for a clearer justification of the timelike branch. The spacelike branch derives its score from a minimum point-to-section distance within the admissible domain. A direct extension of this construction to timelike normals encounters two difficulties: the normal foot need not exist for every input, and the indefinite metric does not provide the same minimum-distance interpretation. Specifically, let R=−β, let nk be a unit timelike normal satisfying ⟨nk,nk⟩Q=−1, and define z=−⟨p,nk⟩Q. The component orthogonal to the normal is
\qquad \langle p_\perp,p_\perp\rangle_{\mathcal Q}=-R^2+z^2.$$ When $|z|<R$, this component is timelike and can be rescaled to obtain a normal foot $q^\star$ on the decision section. When $|z|\geq R$, it is null or spacelike, so this normalization cannot produce a point on the pseudo-hyperboloid. Moreover, even where the normal foot exists, a local timelike normal parameter does not establish a global minimum over the decision section. We therefore retain the timelike branch through an explicitly defined surrogate score. - **Classification rationale** The surrogate uses the signed normal coordinate $z$, whose zero defines the class decision section and whose sign distinguishes its two sides: $$u_k^t(p)=\|a_k\|_2F_R(z), \qquad F_R(z)=R\operatorname{asinh}(z/R).$$ This transformation preserves the zero and sign of $z$. It is also smooth and strictly increasing for every real $z$, since $$F_R'(z)=\frac{1}{\sqrt{1+(z/R)^2}}>0.$$ Thus, it preserves the ordering of each class’s signed normal coordinate and provides a finite score for every finite $z$. The surrogate also has a local geometric interpretation. Where $|z|<R$, the principal timelike normal geodesic can be written as $$p=q^\star\cos(r/R)+Rn_k\sin(r/R),$$ giving $z=R\sin(r/R)$ and $r=R\arcsin(z/R)$. Near the decision section, where $z/R$ is small, the local expansions are $$R\arcsin(z/R)=z+O(z^3/R^2), \qquad F_R(z)=z+O(z^3/R^2).$$ The surrogate therefore agrees with the local signed normal parameter to first order, while extending smoothly beyond the domain where that parameter is available. - **Shared flat limit** The timelike surrogate and the spacelike score use the same scalar transformation: $$u_k^t(p)=\|a_k\|_2F_R(z_k^t(p)), \qquad u_k^s(p)=\|a_k\|_2F_R(z_k^s(p)).$$ For a fixed, bounded normal coordinate $z$, we obtain the following expansion as $\beta\to-\infty$, or equivalently $R\to\infty$: $$F_R(z)=z-\frac{z^3}{6R^2}+O(R^{-4}) \longrightarrow z.$$ Both branches therefore approach the same linear transformation of their respective normal coordinates, which gives them a consistent limiting score scale for adaptive fusion.