- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper introduces SPD Ricci Flow, a graph representation learning framework in which each node’s SPD representation also serves as its local feature-space metric. The method constructs a matrix-valued Ricci-type curvature operator by aggregating discrepancies between neighboring SPD metrics under the AIRM, and uses the resulting metric evolution to update node representations. Experiments on node classification, molecular property prediction, and financial spillover prediction report competitive performance, with ablation studies examining the contributions of graph propagation, adaptive weighting, and the curvature update. Results are competitive across node classification, molecular property prediction, and financial spillover prediction, with ablations testing graph propagation, adaptive weighting, and curvature updates.
Critical strengths
- Theorem 3.2 establishes that the Dirichlet-type energy is nonincreasing along the continuous-time SPD flow and characterizes stationary points through zero Ricci-type curvature. This gives a precise account of the idealized metric evolution and is the theoretical contribution.
- The experimental evaluation covers node classification, molecular property prediction, and financial spillover prediction. In the molecular experiments, retaining the SPDSheaf architecture while replacing its propagation layers helps isolate the update rule. Table 4 reports the highest mean ROC-AUC on four of seven MoleculeNet benchmarks and a tie on Tox21, providing empirical support beyond a single task family.
Critical weaknesses
- Theorem 3.2 proves energy dissipation for a continuous-time flow, but the implemented layer in Eq. (15) adds a finite Euler step, a learnable BiMap and its inverse, and a spectral activation. Positive definiteness of the output alone does not establish energy dissipation. The paper should distinguish properties proved for the idealized flow from those claimed for the trained layer. If monotonicity is intended to explain the latter, sufficient conditions or measured energy trajectories across layers are needed.
- The ablations in Tables 2 and 6 compare the proposed curvature update with an alternative update only under uniform weighting. They do not include a comparison that retains adaptive weights while changing only the update rule. In Table 6, the alternative update achieves higher mean scores than the proposed curvature update on Wisconsin (73.60 versus 69.40) and BACE (88.89 versus 87.71). The existing ablations do not sufficiently establish the additional benefit of the proposed curvature update when combined with adaptive weighting in the full model.
Questions for authors
- Does the energy-dissipation property established in Theorem 3.2 extend to the propagation layer in Eq. (15)? If so, please specify the required conditions, particularly those concerning the step size and spectral activation.
- Can the authors compare the proposed curvature update with the alternative update used in the ablation study while retaining the same adaptive weighting rule and keeping the remaining model and training settings consistent? Please report the same task metrics and variability for both versions.
- Section 4.3 requires the learnable BiMap matrix to be invertible, and Eq. (15) explicitly uses its inverse. How is invertibility maintained throughout training? Please specify the matrix parameterization or constraints used in the experiments, and explain how numerical instability is handled if the matrix becomes nearly singular.
3 comments
W2, Q2:
We thank the reviewer for identifying this missing comparison. We agree that the uniform-weight ablations alone did not establish the additional benefit of the proposed curvature update when combined with adaptive weighting.
We therefore compare the two update rules while retaining the adaptive weighting rule , where is the squared AIRM distance. The encoder, decoder, BiMap parameterization, spectral activation, propagation depth, seeds, and training protocol are kept unchanged. Only the update rule is replaced:
- Adaptive + : , corresponding to the alternative update used in the original ablation.
- Adaptive + Ricci: , corresponding to the full model.
Both variants compute using Eq. (5), with the same adaptive weighting rule evaluated on their respective current SPD states.
Table A. Update-rule comparison under adaptive weighting. Scores are multiplied by 100 and reported as mean ± standard deviation over seeds. The final column reports the difference in means (Ricci minus ).
| Dataset / metric | Adaptive + | Adaptive + Ricci | Difference |
|---|---|---|---|
| Wisconsin: Accuracy | 88.40 ± 4.40 | 89.80 ± 3.52 | +1.40 |
| DBLP: Accuracy | 85.88 ± 0.43 | 85.90 ± 0.39 | +0.02 |
| BACE: ROC-AUC | 87.77 ± 1.70 | 89.70 ± 1.00 | +1.93 |
| SIDER: ROC-AUC | 84.85 ± 0.40 | 85.00 ± 0.30 | +0.15 |
| Financial: Macro-AUPRC | 33.09 ± 0.24 | 35.24 ± 1.01 | +2.15 |
| Financial: NDCG@5 | 29.77 ± 0.24 | 32.67 ± 1.28 | +2.90 |
| Financial: Precision@5 | 12.46 ± 0.11 | 13.57 ± 0.40 | +1.11 |
| Financial: Hit@5 | 62.15 ± 0.35 | 66.11 ± 1.66 | +3.96 |
With adaptive weighting retained, the Ricci update achieves higher mean scores in all eight comparisons. In particular, the ordering highlighted by the reviewer under uniform weighting on Wisconsin and BACE is reversed: the Ricci update now improves their mean scores by 1.40 and 1.93 points, respectively. On the financial task, it improves all four metrics, including Macro-AUPRC by 2.15 points and NDCG@5 by 2.90 points. The differences on DBLP and SIDER are small and should be interpreted cautiously.
These matched comparisons provide evidence that the metric-coupled update contributes beyond adaptive weighting alone, with the clearest mean gains on BACE and the financial task. The contrasting results under uniform weighting also show that its benefit depends on the weighting scheme and dataset, rather than implying uniform superiority of the Ricci update in every setting. We will incorporate these comparisons into the revised Tables 2 and 6 and clarify the corresponding conclusions.
Q3:
We thank the reviewer for requesting this clarification. The invertibility required by Eq. (15) is enforced at every forward pass, rather than assumed from identity initialization. Specifically, each layer learns a matrix , initialized to zero, and constructs , with elementwise . Before applying the BiMaps, we check its smallest singular value. If , the matrix is used directly. Otherwise, for , we replace it with . Both the forward BiMap and the inverse in Eq. (15) use this same effective matrix.
This explicitly handles the near-singular case raised by the reviewer: even if the raw matrix approaches singularity during optimization, its unchecked inverse is never used. The effective matrix satisfies and . Furthermore, bounds the raw matrix norm by , and singular-value flooring preserves this upper bound. Hence , which equals for . These constraints control the amplification associated with the inverse BiMap throughout training, not only at initialization. We have clarified the parameterization, the correction procedure, and these conditioning bounds in Appendix A.4 of the revised manuscript.
I appreciate the authors’ response and clarifications. Having considered the rebuttal, I maintain my original assessment and score.
W1,Q1:
We thank reviewer CjhZ for highlighting the need to link continuous-time energy dissipation to the finite-step propagation layer. We have added necessary and sufficient dissipation conditions and relate them to our experimental implementation (Appendix A.11). Theorem 3.2's dissipation guarantee extends to the layer under step-size conditions. We give two conditions: the activation condition is necessary and sufficient for safe step sizes to exist, while the step-size condition is sufficient for finite-step dissipation.
Consider our model energy Eβ for SPD matrices without assuming commutativity. Fix an undirected graph with symmetric nonnegative weights and at least one positive-weight edge. Assume the layer uses an invertible BiMap shared across all nodes, its tied inverse, and a continuous positive-valued spectral activation ϕ.
1. Activation condition. Assume that, for all x,y∈R, ∣logϕ(ex)−logϕ(ey)∣≤∣x−y∣. Then for every input SPD field G and every shared invertible BiMap W, there exists η0(G,W)>0 such that, for all 0<η<η0(G,W), the finite Euler step preserves positive definiteness and the complete layer satisfies Eβ(G+)≤Eβ(G).
The activation condition requires that the spectral activation not amplify geometric differences between node representations. More precisely, the scalar condition above is equivalent to dA(σϕ(P),σϕ(Q))≤dA(P,Q) for all SPD matrices P,Q, where dA denotes the AIRM distance. The eigenvalue-clipping activation used in our implementation of Eq. (15), ϕ(λ)=max(λ,δ) with δ>0, satisfies this condition because logϕ(ex)=max(x,logδ) does not expand distances on the logarithmic scale. Thus, this activation not only ensures positive-definite outputs but also does not increase the geometric energy used in our model. The condition is not restricted to this activation: another admissible example is scaled clipping, ϕ(λ)=max(cλ,δ) with c>0.
Next, we control the energy change in the Euler update itself: an initially energy-decreasing direction need not decrease energy for an arbitrarily large step. Under the same graph, BiMap, and activation assumptions, let Zi=W⊤GiW be the state after the BiMap and Vi=−RiciSPD(Z) the update direction computed at Z. Holding this direction fixed, define F(s)=Eβ(Z+sV) as the energy along the Euler update path, where s is a trial step size. Thus, F(0) is the energy before the Euler update and F(η) is the energy after the proposal Zi′=Zi+ηVi. We write D=−F′(0) for the initial rate of energy decrease; Theorem 3.2 guarantees D≥0.
2. Step-size condition. Suppose a step size η>0 and a constant M≥0, bounding the second derivative of this energy profile, satisfy Zi+ηVi≻0,∀i;F′′(s)≤M≤η2D,∀s∈[0,η]. Then the output G+ of the complete layer satisfies Eβ(G+)≤Eβ(G).
Under these condition, the layer does not increase energy:
(i) Energy nonincrease in the Euler update. Under Condition 2, the Euler proposals Zi′=Zi+ηVi remain positive definite, and the finite-step second-order error does not exceed the first-order energy decrease, giving Eβ(Z′)≤Eβ(Z).
(ii) Energy nonincrease during activation. Condition 1 ensures AIRM nonexpansiveness of the spectral map, even for noncommuting SPD matrices, hence Eβ(σϕ(Z′))≤Eβ(Z′). Thus, Conditions 2 and 1 control the energy changes from Z to Z′ and from Z′ to σϕ(Z′), respectively. The shared BiMap and its tied inverse preserve AIRM distances and hence energy, so Eβ(G+)=Eβ(σϕ(Z′))≤Eβ(Z′)≤Eβ(Z)=Eβ(G). (iii) Existence of safe small steps. Condition 1 is necessary and sufficient for every input SPD field G and every shared invertible BiMap W to admit sufficiently small safe steps.
Our model's shared invertible BiMap and tied inverse meet the theorem's requirements. The BiMap need not be orthogonal or close to the identity. We take the spherical experiment case (Fig.2) as an easy example to show an direct substitution of its parameters into the bound: gi=e2uiI2,β=0.5,η=0.01,ϕ(λ)=max(λ,10−9). This activation satisfies the condition above. For this conformal setting with maximum weighted degree d_\max^w, we can prove that every update preserves positive definiteness and does not increase energy when η≤\frac1{d_\max^w(1+1/\sqrt{2eβd})}. When the actual mesh has maximum degree 6, setting d=2 and β=0.5 gives η≤0.11664. The experimental step size η=0.01 lies within this guaranteed range. Thus, these conditions directly relate to the activations and step sizes used in our experiments. This explicit bound applies to the conformal spherical experiment; the preceding necessary-and-sufficient criterion and finite-step conditions cover general noncommuting SPD learning layers.