REVIEW 3 major objections 5 minor 14 references
Enhancing Graph Representation Learning with Localized Topological Features
T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash
Pith's one-line read Localized loop-counting features from extended persistent homology make graph neural networks strictly more expressive than 3-WL and 4-WL on specific graph pairs, and lift node classification and link prediction accuracy.
desk verdict A solid extension of the authors' own localized-PH line, with genuinely new expressiveness bounds—but the theorems are proved for exact diagrams under a tie-break filter, while the deployed model uses lossy vectorized features, so the theoretical support for the implemented method is looser than the abstract suggests. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the extended persistence diagram (EPD) of a vicinity graph filtered by a distance function. Ordinary persistence records when a loop or component is born but not when it dies, so the paper uses extended persistence, which tracks each structure across both ascending and descending filtrations and captures the whole life span of loops and of the connected component. The EPD is computed on the hop-bounded neighborhood of a node for node classification, and on the intersection of the two endpoints' neighborhoods for link prediction, with distance measured under a graph metric. Because diagrams live in an infinite-dimensional space, the paper vectorizes them into persistence images (Gaussian-smoothed pixel grids) and augments the vector with counts of nodes at each distance layer, intra-layer edges, and crossing edges to form PI+; these vectors then reweight message passing in the Persistence Enhanced Graph Network (PEGN).
What would settle it
Take the Shrikhande and 4×4 Rook graphs and the CFI pair, compute the implemented 25-pixel persistence image with the paper's default Gaussian smoothing plus the PI+ counts, and check whether the feature vectors, and the trained GNN embeddings, are identical up to symmetry; if they coincide, the separation proved for exact EPDs has been lost in vectorization and the implemented method's expressiveness claim fails.
Extended reading notes
Core claim
The paper claims that localized extended persistence diagrams (EPDs) are a rich and provably powerful representation of local graph structure. With the shortest-path distance as the filter function, the EPD of a node's hop-bounded vicinity graph counts the cycles formed by that node's neighbors and other layer-wise connectivity statistics; this lets the features separate the Shrikhande graph from the 4×4 Rook graph, a pair 3-WL cannot separate, and separate a CFI pair (non-isomorphic graphs built to defeat low-dimensional Weisfeiler-Leman tests) that 4-WL cannot separate. The same EPD is simultaneously bounded: with shortest-path-distance filters, 1D EPDs are less powerful than 4-WL overall, and for most pairs of sufficiently large regular graphs of a fixed degree an EPD with a bounded number of hops tells the two graphs apart. The paper further claims that vectorized versions of these features, namely persistence images plus hand-added layer and edge counts (PI+), improve node classification and link prediction beyond strong baselines when used to reweight GNN messages, and that end-to-end learning of the filter function is feasible but mainly pays off on large, dense link-prediction benchmarks.
Load-bearing premise
The weakest load-bearing premise is that the lossy vectorization used in the deployed model, namely persistence images plus hand-added counts, preserves the distinguishing power that the paper proves for exact persistence diagrams; the paper does not prove this transfer.
Editorial extensions
If this is right
- Augmented GNNs can tell apart the Shrikhande graph and the 4×4 Rook graph, so their distinguishing power on that pair exceeds 3-WL.
- A CFI pair that 4-WL cannot separate is separated by the same localized EPD construction, although Theorem 3 keeps 1D EPDs below 4-WL overall.
- PI+ restores layer and edge counts that plain persistence images blur, giving node and edge embeddings access to local structural statistics that standard message passing cannot easily recover.
- End-to-end learning of localized topological features is computationally heavy; the benchmarks show it helps mainly on large and dense link-prediction graphs, while precomputed curvature-based filters usually win for node classification.
Reading between the lines
- The expressiveness theorems are proved for exact EPDs with shortest-path-distance filters, while the deployed model uses lossy persistence images and an Ollivier-Ricci-curvature-based graph metric; a direct comparison on the Theorem 1 and Theorem 2 counterexample pairs would show whether the vectorized feature retains the proved separation power.
- The PI+ design points to a general recipe: pair any lossy topological vectorization with hand-added structural counts that the vectorization is known to blur, and the same pattern could be extended to higher-order recoverable statistics such as numbers of short cycles through a node.
- A cheap ablation replacing Gaussian persistence images with exact-diagram-derived counts or Dirac peaks would isolate whether the empirical gains come from the smoothed persistence image or from the explicitly appended structural counts, which the paper does not run.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a family of localized topological features for graph representation learning, based on extended persistent homology (EPH) computed on k-hop vicinity graphs of nodes (for node classification) and on intersections of k-hop neighborhoods (for link prediction). The diagrams are vectorized as persistence images (PI) and an augmented version (PI+) that concatenates level counts and intra/inter-layer edge counts. These features are combined with GNN backbones in the proposed Persistence Enhanced Graph Network (PEGN), with variants that either combine multiple precomputed filter functions or learn the filter function end-to-end. The paper reports accuracy and AUC improvements over several baselines on standard node classification and link prediction benchmarks. It also states four theorems (Theorems 1-4) about the expressiveness of exact extended persistence diagrams relative to k-WL, including examples where EPDs separate pairs that 3-WL or 4-WL cannot distinguish, an upper bound showing 1D EPD is less powerful than 4-WL, and a result on distinguishing most pairs of regular graphs.
Significance. If the expressiveness results were proved for the features actually fed into the GNN, this would be a meaningful contribution integrating topological data analysis with GNNs, with concrete guidance for practitioners. The empirical study is broad: it covers seven node-classification and five link-prediction datasets, compares against a wide range of strong baselines, reports mean and standard deviation over ten runs, and releases source code. The counterexample-based separation results for exact EPDs are interesting and directly relevant to the TDA-for-graphs literature. The main weakness is that the theoretical claims are established only for exact persistence diagrams under a hand-crafted tie-break filter, while the implemented model consumes persistence-image vectors computed with standard lower/upper-star filtrations; no theorem bridges this gap. The end-to-end learning variants are honestly evaluated, including their computational cost, which is a strength of the paper.
major comments (3)
- [Section 4.1 (Eq. (2)) and Section 4.2] Theorems 1 and 2 are proved for exact extended persistence diagrams computed with the tie-break filter function in Eq. (2), namely f(uv)=max(f(u),f(v)) when endpoint levels differ and f(u)+0.5 when they are equal (and the analogous descending rule in Eq. (3)). The deployed PEGN model in Section 4.2 instead uses standard lower-star and upper-star extensions of shortest-path or Ollivier-Ricci distance filters, and then compresses the diagrams into persistence images (or PI+ vectors, Eq. (4)). The footnote to Eq. (2) and Appendix D report only that node-classification accuracies under the two filters are nearly identical on the benchmark datasets; they do not establish that the specific separating pairs from Theorems 1 and 2 (Shrikhande vs. 4x4 Rook, and the G(0)/G(1) pair) are still separated by the persistence-image vectorization used in the model, with the stated resolution and sigma parameters. Since the abstract and introduction use the expressiveness results to justify the proposed framework, this gap is load-bearing: the proven statements concern an object that the model does not consume. The authors should either prove that the deployed features preserve the separation for these pairs, or explicitly restrict the theoretical contribution to exact EPDs and adjust the corresponding claims about the implemented method.
- [Appendix B (Theorem 3)] The proof of Theorem 3 relies on the assertion that 'there exists a bijective hashing function HASH that can approximate the computation of the 1D EPDs' (Eqs. (13)-(14)). This is not a formal simulation argument: Algorithm 1 is a union-find procedure whose state includes the evolving disjoint-set partition and the active cycle representatives C_i, and the paper does not show that the subgraph-WL coloring sequence injectively encodes this state at every iteration. Moreover, the word 'approximate' is too weak for an upper-bound theorem, because an approximate simulation could in principle merge persistence points that the exact algorithm distinguishes. Consequently, the conclusion that 1D EPD is 'less powerful than 4-WL' does not follow from the argument as written. A precise reduction is needed, such as showing that each extended persistence point (b,d) is determined by a histogram computed by subgraph-WL, or a direct argument that any graph pair distinguished by the 1D EPD is distinguished by 4-WL.
- [Section 4.1 bullet list and Theorem 4] The paragraph before Theorem 1 claims that adding EPDs as topological position embeddings makes the 'resulting enhanced GNN' more expressive than 2-WL, not less expressive than 3-WL, and less expressive than 4-WL. However, Theorems 1-3 are statements about the distinguishing power of the exact EPD itself, not about the GNN architecture in Eqs. (5)-(6) or about the vectorized PI/PI+ representation that the GNN actually receives. No theorem shows that the message-passing network with PI or PI+ features can recover the relevant persistence points, e.g., the (1.5, 0.5) points used in the proof of Theorem 1, even though the paper asserts these queries 'can be easily implemented using MLP.' Thus the theoretical results justify a property of an idealized feature extractor, not of the implemented architecture. The authors should either add a formal statement about the implemented model (for instance, showing that an MLP on PI+ can compute the layer counts and Betti numbers used in the separation proofs) or clearly delimit the expressiveness claims to exact EPDs.
minor comments (5)
- [Abstract and Section 5.1] The abstract and conclusion state that the method 'achieve[s] state-of-the-art performance,' but Table 2 shows that PEGN does not beat GAT on Cora (82.7 vs. 83.0) or Citeseer (71.9 vs. 72.5), and Table 3 shows HGCN outperforms PEGN on Citeseer link prediction (96.6 vs. 95.1). The claim should be qualified as state-of-the-art on most benchmarks.
- [Section 1] The introduction says 'We propose the first method to adapt persistent homology for localized graph representation,' then states that the work builds on two preliminary conference papers that were 'the first to adapt persistent homology for localized node-level and edge-level representation learning.' This phrasing should be reconciled to avoid an apparent novelty contradiction.
- [Section 4.2.2] In Eq. (4) and the surrounding text, the counts n_j, n_{j,j}, and n_{j,(j+1)} are defined for the node vicinity graph G^k_u, but in the link prediction setting the vicinity graph is the intersection G^k_{uv}. The notation should be adapted or clarified so that PI+ is unambiguously defined for the pair-wise setting.
- [Appendix A] The proof of Theorem 2 relies on the cited result that the constructed graphs G(0) and G(1) 'cannot be differentiated by 4-WL (Yan et al., 2024b).' Since this is the sole external load-bearing result for the 4-WL part of the theorem, the reference should include the specific theorem or proposition being invoked, so that the reader can verify the claim without recovering it from a broader paper.
- [Section 3.2 and Theorem 1 proof] The notation for persistence points is inconsistent: Section 3.2 states that for loops the death time is smaller than the birth time, while the proof of Theorem 1 refers to 'persistence points with persistence value (1.5, 0.5)' without specifying whether the pair is (birth, death) or (death, birth). Please standardize the order and define it once in Section 3.2.
Circularity Check
No significant circularity: the theoretical results are genuine invariant computations, and the gap between exact EPDs and deployed persistence images is an unproven transfer, not a circular reduction.
full rationale
The paper's derivation chain is not circular. Theorems 1-4 prove properties of extended persistence diagrams computed from explicit filter functions; the proofs construct concrete graph pairs (Shrikhande vs. 4x4 Rook, CFI-derived pair) and exhibit persistence-point counts, such as the number of 1D points at (1.5, 0.5) equaling the Betti number of the layer-1 induced subgraph under the tie-break filter of Equation (2). These are invariant computations, not definitions of the target in terms of the input. The deployed PEGN uses fixed precomputed persistence images or PI+ and learned MLP weights; no fitted topological parameter is relabeled as a prediction, and the benchmark comparisons are external. The expressiveness theorems are stated for exact EPDs with the tie-break filter, while the model uses persistence images with other filters; the paper explicitly flags this difference in the footnote to Equation (2) and Appendix D. That is an unproven transfer or correctness risk, not a circular reduction. Self-citations to Zhao et al. 2020, Yan et al. 2021, Yan et al. 2022b, and Yan et al. 2024b provide provenance and algorithmic tools, but the load-bearing mathematical facts (CFI graphs resisting 4-WL, subgraph-WL upper bounds) are standard external results also cited from Cai et al. 1992, Huang et al. 2023a, and Frasca et al. 2022. Nothing labeled a prediction is fitted by construction, so the central claim retains independent content.
Assumptions & free parameters
free parameters (7)
- Persistence image resolution =
25 pixels
- Persistence surface Gaussian sigma =
1
- Persistence image weight thresholds =
0 and 1
- Vicinity graph hop k =
1 or 2 depending on dataset
- HKS diffusion parameter t =
0.1 and 10
- Ollivier-Ricci curvature offset =
+1
- Tie-breaking offset in proof filters =
0.5
assumptions (7)
- standard math Persistent homology and extended persistence definitions and stability results from Cohen-Steiner et al. 2007, 2009 and Adams et al. 2017
- standard math The 4x4 Rook graph and Shrikhande graph are not distinguishable by 3-WL
- standard math The constructed CFI pair G(0) and G(1) are non-isomorphic and 4-WL indistinguishable
- standard math Subgraph WL rooted on 2-tuples is less powerful than 4-WL
- ad hoc to paper A bijective HASH function can simulate the union-find EPD computation within subgraph WL
- domain assumption Ollivier-Ricci curvature with alpha=0.5 and Sinkhorn approximation is a suitable graph metric for filtering
- standard math Random regular graph estimates from Bollobas 1980 and Feng et al. 2022
Cite this review
Pith. "Pith review of Enhancing Graph Representation Learning with Localized Topological Features." pith.science (2026). https://pith.science/paper/MYRSXSKN
@misc{pith2026250109178,
author = {Pith},
title = {Pith review of: Enhancing Graph Representation Learning with Localized Topological Features},
year = {2026},
howpublished = {\url{https://pith.science/paper/MYRSXSKN}},
note = {Machine review of arXiv:2501.09178}
}
read the original abstract
Representation learning on graphs is a fundamental problem that can be crucial in various tasks. Graph neural networks, the dominant approach for graph representation learning, are limited in their representation power. Therefore, it can be beneficial to explicitly extract and incorporate high-order topological and geometric information into these models. In this paper, we propose a principled approach to extract the rich connectivity information of graphs based on the theory of persistent homology. Our method utilizes the topological features to enhance the representation learning of graph neural networks and achieve state-of-the-art performance on various node classification and link prediction benchmarks. We also explore the option of end-to-end learning of the topological features, i.e., treating topological computation as a differentiable operator during learning. Our theoretical analysis and empirical study provide insights and potential guidelines for employing topological features in graph learning tasks.
Figures
Figures from the paper (1 more)
Reference graph
Works this paper leans on
-
[6]
Boosting the cycle count- ing power of graph neural networks with i $ˆ2$-GNNs
Yinan Huang, Xingang Peng, Jianzhu Ma, and Muhan Zhang. Boosting the cycle count- ing power of graph neural networks with i $ˆ2$-GNNs. In The Eleventh International Conference on Learning Representations, 2023a. Zexi Huang, Mert Kosan, Arlei Silva, and Ambuj Singh. Link prediction without graph neural networks. arXiv preprint arXiv:2305.13656 , 2023b. Imm...
-
[9]
Chien-Chun Ni, Yu-Yao Lin, Jie Gao, and Xianfeng Gu
Accessed: 2021-10-01. Chien-Chun Ni, Yu-Yao Lin, Jie Gao, and Xianfeng Gu. Network alignment by discrete ollivier-ricci flow. In International Symposium on Graph Drawing and Network Visualiza- tion, pages 447–462. Springer,
work page 2021
-
[13]
A limit theorem for the $1$st Betti number of layer-$1$ subgraphs in random graphs
arXiv preprint. Available at arXiv:1911.00585. Jake Topping, Francesco Di Giovanni, Benjamin Paul Chamberlain, Xiaowen Dong, and Michael M Bronstein. Understanding over-squashing and bottlenecks on graphs via curvature. In International Conference on Learning Representations ,
work page Pith review arXiv 1911
-
[14]
Topogan: A topology-aware generative adversarial network
34 Enhancing Graph Representation Learning with Localized Topological Features Fan Wang, Huidong Liu, Dimitris Samaras, and Chao Chen. Topogan: A topology-aware generative adversarial network. In Computer Vision–ECCV 2020: 16th European Confer- ence, Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16 , pages 118–136. Springer,
work page 2020
-
[1980]
doi: 10.1016/S0195-6698(80)80030-8
ISSN 0195-6698. doi: 10.1016/S0195-6698(80)80030-8. B´ ela Bollob´ as. Distinguishing vertices of random graphs. In B´ ela Bollob´ as, editor,Graph Theory, volume 62 of North-Holland Mathematics Studies , pages 33–49. North-Holland,
-
[1982]
Jin-Yi Cai, Martin F¨ urer, and Neil Immerman
doi: https://doi.org/10.1016/S0304-0208(08)73545-X. Jin-Yi Cai, Martin F¨ urer, and Neil Immerman. An optimal lower bound on the number of variables for graph identification. Combinatorica, 12(4):389–410,
-
[2008]
Pitfalls of graph neural network evaluation
Oleksandr Shchur, Maximilian Mumme, Aleksandar Bojchevski, and Stephan G¨ unnemann. Pitfalls of graph neural network evaluation. arXiv preprint arXiv:1811.05868 ,
-
[2010]
Persistence weighted gaussian kernel for topological data analysis
Genki Kusano, Yasuaki Hiraoka, and Kenji Fukumizu. Persistence weighted gaussian kernel for topological data analysis. In International Conference on Machine Learning , pages 2004–2013. PMLR,
work page 2004
Show all 14 references
-
[2013]
Ripsnet: a general architecture for fast and robust estimation of the persistent homology of point clouds
Thibault de Surrel, Felix Hensel, Mathieu Carri` ere, Th´ eo Lacombe, Yuichi Ike, Hiroaki Kurihara, Marc Glisse, and Frederic Chazal. Ripsnet: a general architecture for fast and robust estimation of the persistent homology of point clouds. In ICLR 2022 Workshop on Geometrical...
2022
-
[2016]
Graph u-nets
Hongyang Gao and Shuiwang Ji. Graph u-nets. In international conference on machine learning, pages 2083–2092. PMLR,
-
[2018]
Wasserstein stability for persistence diagrams
Primoz Skraba and Katharine Turner. Wasserstein stability for persistence diagrams. arXiv preprint arXiv:2006.16824,
2006 arXiv
-
[2019]
The density of expected persistence diagrams and its kernel based estimation
Fr´ ed´ eric Chazal and Vincent Divol. The density of expected persistence diagrams and its kernel based estimation. arXiv preprint arXiv:1802.10457 ,
-
[2020]
Notes on an elementary proof for the stability of persistence diagrams
Primoz Skraba and Katharine Turner. Notes on an elementary proof for the stability of persistence diagrams. arXiv preprint arXiv:2103.10723 ,
-
[2021]
Christopher Morris, Martin Ritzert, Matthias Fey, William L Hamilton, Jan Eric Lenssen, Gaurav Rattan, and Martin Grohe
Accessed: 2021-10-01. Christopher Morris, Martin Ritzert, Matthias Fey, William L Hamilton, Jan Eric Lenssen, Gaurav Rattan, and Martin Grohe. Weisfeiler and leman go neural: Higher-order graph neural networks. In Proceedings of the AAAI Conference on Artificial Intelligence ,...
2021
Reviewed August 10, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.