- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 1
Read review
Paper summary
The paper introduces Dual-Order Rényi–Dirichlet Graph Autoencoders (RD-GAEs), a graph encoder for anomaly detection on simplex-valued data. It learns representation based on the probability simplex of the input. It combines stochastic graph operators, stochastic category mapping, gated probability-temperature map, and Dirichlet bottleneck. The model uses a Rényi order below 1 for training, to be exact. After training, it uses an order above one to calculate anomaly scores. The paper gives conditional score-separation results and contamination bounds. Experiments cover two image datasets and two biological datasets.
Critical strengths
- The reported improvements are substantial, particularly for AUPRC.
Critical weaknesses
- The writing is poor and lacks of clarity. Sections 4.3 and 4.4 repeat the same discussion. Implementation details need more clarification to ensure reproducibility.
- The novelty of the proposed method is not well established. The combination of the components can be valuable, but the paper needs to clarify what new insights it provides and how it can be used to solve practical problem. Its practical value also remains uncertain because the evaluation uses only synthetic perturbations.
Questions for authors
- Does low-order training still improve performance when a KL-trained model uses the same Rényi detection-order selection procedure?
- The main text mentions the detection order is selected by validation AUROC. However, the text below Appendix Table 3 uses AP first, then AUROC to break ties. Which criterion was used in the implementation?
- Provide more justification on the results in Figure 4. What are the actual Omniglot contamination rate?
- What formula did the implementation use for the softmax gate ablation? Table 6 omits the coefficient included in the main text.
- Did the softmax-gate ablation use a suitable temperature range?
- Can you provide the complete edge-weight formula described in Section 2.1? How do you define the receiver’s local consensus and set the strength of the downweighting?
- Theorems 3.1 and 3.3 concern fixed parameters and graph operators. Do you have results that compare low-order and KL training at different contamination rates, with the detection procedure held fixed?
- What are the limitations of the proposed methods?
- What is the scalability of the RD-GAEs?
- Appendix Table 3 states that RD-GAE uses early stopping, while the baselines run to the epoch cap. How were baseline checkpoints selected?
- What is ILR-transformated Input?
6 comments
Response to Reviewer bNkE (part 1.5/2), question 1 -10)
Q1
In additional experiments, following your suggestion, both models use the same validation-based Rényi detection-order procedure. KL training obtains AUROC/AUPRC , whereas subunit-order training obtains . The modest gain supports a training-order contribution but does not establish its necessity. See revised Appendix E.1, Table 6.
Q2
We thank the reviewer for identifying this inconsistency. We audited and corrected the text: the implementation uses validation AP, reported as AUPRC, as the primary criterion, AUROC as the first tie-breaker, and the smaller order as the final tie-breaker. See the remark following Theorem 2.4 in revised main Section 2.4 and revised Appendix Table 3.
Q3
We verified that Omniglot contains contaminated images among (), exactly two of 20 images per character; the target fraction is in each evaluated partition. We also expanded the qualitative interpretation of the reconstructions. See revised main Figure 4 and revised Appendix Table 2.
Q4
We corrected the ablation formula to ; the omitted coefficient was a typographical error in the table, not in the implementation. See revised Appendix Table 8; the primary escort gate is defined in revised main Equation (4).
Q5
We thank the reviewer for raising this design issue. Following your suggestion, we clarify that the reported softmax-gate ablation uses the RD-GAE-selected so that only the gate formula changes. This is a matched ablation, not an independent temperature search for the softmax gate; we have therefore limited its interpretation and do not claim that it is the globally best softmax-gate result. The searched RD-GAE temperature range is reported in revised Appendix Table 3, and the matched ablation shown in Table 8.
Q6
We have added the full formula to revised main Section 2, Equation (2). For the row-stochastic base operator , the receiver consensus, sender disagreement, and normalized robust edge weight are
d_{ij}^{(\ell)}=\tfrac12\|\bm h_j^{(\ell)}-\bm c_i^{(\ell)}\|_1,\qquad \widetilde P_{ij}^{(\ell)}= \frac{P_{ij}\exp\{-\rho_gd_{ij}^{(\ell)}\}} {\sum_kP_{ik}\exp\{-\rho_gd_{ik}^{(\ell)}\}}.$$ Here $\rho_g$ controls the downweighting strength; we use $\rho_g=5$, while $\rho_g=0$ recovers $\bm P$. Self-loops enter the consensus and normalization. In the primary encoder-only model, these weights are recomputed at each encoder layer, while decoder layers use the identity graph operator. ## Q7 In updated experiments, we added a paired MNIST experiment comparing subunit-order and KL training at $0\%,5\%,10\%$, and $20\%$ training contamination, while fixing validation/test contamination at $10\%$ and scoring every model at $\lambda_{\mathrm{det}}=1.5$. As shown in revised Appendix Figure 12, subunit-order training improves AUROC at every rate and improves AUPRC at all nonzero contamination rates, with the AUPRC gain increasing to $0.022$ at $20\%$. At $0\%$, KL training has a marginally higher AUPRC. These results isolate the training order and support its benefit specifically when the training data are contaminated. ## Q8 Following your suggestion, the revised main Section 5 now identifies four limitations: the theory gives sufficient rather than universal guarantees; controlled contamination remains necessary where definitive labels are unavailable; repeated row-stochastic category mappings may attenuate profile differences, while overly strong escort sharpening may remove information; and the moderate tonsil AUROC suggests that stronger semantic representations are needed for biologically heterogeneous minority populations. It also describes MNIST and Omniglot as controlled total-mass-normalized image benchmarks rather than naturally compositional data. ## Q9 We added time and memory complexity in revised Appendix C.5. For a sparse fixed-$k$ graph, RD-GAE message passing scales linearly with the number of nodes; the main large-sample limitation of the current implementation is the $O(N^2K)$ exact graph-construction step. The revised manuscript states: > “A sparse RD-GAE forward–backward pass has time complexity $O\{S\sum_\ell(|E|d_\ell+Nd_\ell d_{\ell+1})\}$ and activation-memory complexity $O(SN\sum_\ell d_\ell)$. Thus, for a fixed-$k$ graph, $|E|=O(kN)$ and message passing is linear in $N$.” ## Q10 We clarified that each baseline uses its final Stage 3 checkpoint at the 200-epoch cap; no validation-based checkpoint is selected for the baselines. See revised Appendix Table 4 and the paragraph immediately following it.Response to Reviewer bNkE (part 2/2, question 11, ethics concern)
Q11
We now state that the isometric log-ratio (ILR) transform maps a -part composition to orthonormal log-ratio coordinates using a selected orthonormal basis (Aitchison, 1982). The ILR baselines use mean-squared reconstruction error in that coordinate space. See revised Appendix E.1, particularly the paragraph preceding Table 6.
Ethics concerns
Q1
Figure 1 is developed using LaTeX TikZ with caption text polished using AI (disclosed AI use for writing).
Q2
Thank you for the suggestion. We address this concern under Weakness 1.
Response to Reviewer bNkE (part 1/2)
We thank the reviewer for identifying these gaps and limitations. Following the your suggestions, we clarified the methodology, conducted additional analyses, and revised the main manuscript and Appendix accordingly (see the colored text).
Weakness 1
We thank the reviewer for helping us improve the presentation, and we apologize for the duplicated discussion. Following the reviewer’s suggestion, we removed the repeated section and expanded the reproducibility details. Revised Appendix Tables 2, 3, 4, and 5 now report, respectively, the application architecture and graphs, randomized search space, stage-wise budgets and selection, and contamination operators. The accompanying GitHub repository will provide preprocessing code, configurations, seeds, and scripts for all three stages.
Weakness 2
Following your suggestion, the final paragraph of the revised main Introduction now states three contributions: end-to-end simplex preservation, separate Rényi orders for fitting and detection, and theory for attenuation, score separation, and graph-mediated contamination. We also added an unperturbed tonsil case study with annotation-defined minority cell populations, rather than claiming independently verified pathological anomalies. See the last paragraph of revised main Section 1 and revised main Section 4.7, especially Figures 6 and 7.