Pith. sign in

REVIEW 4 major objections 7 minor 64 references

Enhancing Unsupervised Graph Few-shot Learning via Set Functions and Optimal Transport

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read Graph few-shot learning can be done without base-class labels, and STAR reports the best accuracy among listed baselines on nearly all tested few-shot settings across eight graphs.

desk verdict Strong empirical results and a sensible set-level contrastive idea, but the OT calibration step is dimensionally wrong as written and the main corollary proof has an inequality direction error—needs fixing before the claims hold. read the letter →

arxiv 2501.05635 v1 pith:6PPOYUG6 submitted 2025-01-10 cs.LG

classification cs.LG
keywords unsupervisedgraphfew-shotlearningsetfunctionsoptimaltransportcontrastivenodeclassificationpermutation-invariantrepresentationsdistributioncalibration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes STAR, a method for few-shot node classification that does not need any class labels during meta-training. It claims that combining instance-level contrastive learning with set-level features, obtained by retrieving a node's most similar neighbors and encoding them with a permutation-invariant set function, captures task-relevant structure that instance-level features alone miss. It further claims that moving the small support set toward the query-set distribution with optimal transport reduces the distribution shift that hurts few-shot classifiers. On eight benchmark graphs, STAR reports the highest accuracy among all listed baselines in nearly every few-shot setting, including the two large-scale graphs. The theory argues that the concatenated representation carries strictly more task-relevant mutual information than either feature type alone and has a smaller Bayes-error and generalization-gap upper bound.

What carries the argument

The load-bearing object is the neural set function $\Psi(\Omega) = \mathrm{MLP}(\mathrm{sumpool}(\Omega))$, a permutation-invariant map from an unordered collection of neighbor embeddings to one vector, applied to the top-$k$ nodes retrieved by dot-product similarity between two augmented views. The second load-bearing object is the entropy-regularized optimal transport plan $\lambda^* = \arg\min_{\lambda \in \Lambda} \langle \lambda, D \rangle - \epsilon H(\lambda)$, solved by the Sinkhorn-Knopp algorithm, where $D$ is the pairwise distance matrix between support embeddings $Z_{\mathrm{spt}}$ and query embeddings $Z_{\mathrm{qry}}$. The transported support embeddings $\hat{Z}_{\mathrm{spt}} = \lambda^{*\top} Z_{\mathrm{spt}}$ are then used to train the classifier, while the meta-training objective combines the instance-level contrastive loss $\mathcal{L}_{\mathrm{ins}}$ and the set-level contrastive loss $\mathcal{L}_{\mathrm{set}}$.

What would settle it

Run the method on a graph where connected nodes usually belong to different classes; if the label agreement of the top-$k$ retrieved neighbors is no better than chance and STAR's advantage over its variant without the set-level contrastive loss disappears, then the set-level contrastive mechanism is not doing the claimed work. A concrete check is to plot retrieval accuracy during training on such a graph and compare STAR with its no-set-loss ablation.

Watch

Extended reading notes

Core claim

STAR's central claim is that a graph node classifier for unseen classes can be trained without seeing any base-class labels and still outperform supervised meta-learners, provided the representation joins two complementary views of each node: an instance-level view from graph contrastive learning, and a set-level view formed by retrieving the node's top-k most similar nodes from the other augmented view, encoding that neighbor set with a sum-pooling MLP set function, and contrasting the resulting set embeddings. At meta-test time, STAR treats the remaining problem as a distribution shift between the tiny labeled support set and the larger unlabeled query set, and uses entropy-regularized optimal transport (the Sinkhorn-Knopp algorithm) to move support embeddings into the query region before fitting a linear classifier. Empirically, Tables 3-5 show that this pipeline achieves the highest accuracy among graph embedding, traditional meta-learning, and graph meta-learning baselines on CoraFull, Coauthor-CS, Cora, WikiCS, ML, CiteSeer, ogbn-arxiv, and ogbn-products in nearly every few-shot setting. The theoretical section proves that for the concatenated representation $Z = \tilde{H}\,||\,\tilde{S}$, the mutual information with the task satisfies $I(Z; T) \ge \max\{I(\tilde{H}; T), I(\tilde{S}; T)\}$, that the Bayes-risk upper bound satisfies $U(P(Z)) \le \min\{U(P(\tilde{H})), U(P(\tilde{S}))\}$, and that the generalization gap is bounded by a term depending on the weight norm, the rank of the covariance, and the support size.

Load-bearing premise

The load-bearing premise is that the $k$ most similar nodes fetched for a target node from the altered copy of the graph mostly belong to the same class as the target node, so that contrasting the two halves of the retrieved set teaches class-relevant structure rather than noise.

Editorial extensions

If this is right

  • STAR's results imply that abundant unlabeled graph data can replace labeled base-class tasks for few-shot node classification, removing a major annotation bottleneck.
  • The optimal transport calibration step is directly shown to help: the ablation without it drops accuracy across all datasets, so aligning support and query distributions is a transferable fix for few-shot classifiers.
  • Concatenating instance-level and set-level representations is theoretically and empirically better than either alone, supporting the view that graph few-shot episodes should be modeled as sets, not just as collections of independent nodes.
  • The generalization bound suggests a concrete practical rule: during meta-testing, constrain the classifier's weight norm (the parameter $\gamma$) to shrink the gap, and the paper motivates weight regularization for this purpose.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A test the paper leaves implicit: on graphs where connected nodes tend to have different labels, the top-$k$ retrieval premise should be checked by measuring label agreement of retrieved neighbors; if agreement is near chance, set-level contrastive learning is likely learning noise rather than category structure.
  • The optimal transport plan uses uniform marginals, so if support and query sets have different class proportions, an unbalanced transport variant that relaxes those marginals is a natural extension that could calibrate better.
  • Because the set function is a sum-pooling MLP but the paper notes the framework adapts to any set encoder, swapping in an attention-based set encoder would test whether modeling pairwise interactions among retrieved neighbors adds further gains.
  • Theorem 5.3 motivates weight-norm regularization during meta-testing, but the paper does not systematically vary this regularization; a sensitivity study would turn the bound into a practical hyperparameter recommendation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes STAR, an unsupervised graph few-shot learning method for node classification. During meta-training, STAR runs instance-level graph contrastive learning on two augmented views of the unlabeled graph, then constructs set-level positive pairs by retrieving top-k similar nodes across views and feeding them to a permutation-invariant set encoder, optimizing a second contrastive objective. At meta-test time, STAR computes support and query embeddings, solves an entropic optimal transport problem between them, and uses the transport plan to produce a 'transported support set' on which a linear classifier is trained. The authors prove two information-theoretic and generalization statements and report experiments on eight datasets where STAR outperforms graph embedding, traditional meta-learning, and graph meta-learning baselines.

Significance. If the method is exactly as intended, the contribution is potentially significant: it removes the need for base-class labels in graph few-shot learning, a practical bottleneck, and it combines set-level permutation-invariant features with OT-based distribution calibration. The experiments are broad (eight datasets, including two OGB large graphs) and the reported gains are often large. Theorem 5.1 is a correct application of the data processing inequality (concatenation cannot reduce mutual information), and Theorem 5.3 follows a standard Rademacher-complexity argument, so the theoretical scaffolding is not circular. However, two load-bearing pieces of the write-up, the OT calibration formula and the proof of Corollary 5.2, contain errors that prevent the paper, as written, from supporting its central claims. These are fixable but require substantive revision and re-verification.

major comments (4)
  1. [Section 4.3 and Algorithm 1, lines 17-19] The transported support formula is dimensionally inconsistent and cannot be used to train a classifier as stated. With Z_spt in R^{NK x d} and lambda* in R^{NK x NQ}, the definition hat_Z_spt = lambda*^T Z_spt produces a matrix in R^{NQ x d}, one row per query point. Training a linear classifier on these rows requires labels, but the support labels have length NK and query labels are unavailable at inference; each row is a convex mixture of support points of mixed classes, so a label-assignment rule is absent. A barycentric map that preserves support rows and their labels would be hat_Z_spt = D_alpha^{-1} lambda* Z_qry (or an equivalent normalized form), not lambda*^T Z_spt. Because Section 4.3 and Algorithm 1 define the method only through the transposed formula, the accuracies in Tables 3-6 cannot be reproduced from the written method, and the OT ablation in Table 6 may reflect an unspecified or leakage-prone label assignment. The authors must correct the formula, specify the supervised step (including which labels are used), and rerun the experiments.
  2. [Appendix A.2.2, Eq. (18)] The proof of Corollary 5.2 confuses the task variable T with the label variable Y and then asserts the wrong inequality. Eq. (16) establishes I(Z;T) - I(H_tilde;T) = I(S_tilde;T | H_tilde); to derive a statement about H(Y|Z) one must set Y = T or redo the argument with Y everywhere. Even after that substitution, the equations imply H(Y|Z) = H(Y) - I(Y;Z) and H(Y|H_tilde) = H(Y) - I(Y;Z) + I(S_tilde;Y | H_tilde), hence H(Y|Z) <= H(Y|H_tilde). The text instead concludes 'H(Y|Z) >= H(Y|H_tilde)', which is the reverse of what Corollary 5.2 needs. The corollary's conclusion may be recoverable from Theorem 5.1, but the proof as written is inconsistent and must be rewritten.
  3. [Section 4.2 and Fig. 4] The usefulness of the set-level contrastive objective rests on the assumption that the top-k most similar nodes retrieved from the augmented view share the semantic class of the target node, so that splitting them into positive and negative sets learns class-discriminative structure. The only direct evidence is Fig. 4, which reports retrieval accuracy on two datasets under 2-way settings where random retrieval is 50%; the plotted curves do not clearly exceed that chance level, and no retrieval accuracy is reported for the larger datasets or for N > 2. The paper should report retrieval accuracy against the chance baseline across all datasets and settings, or provide a different argument for why retrieved sets are semantically coherent. Without this, the claim in Theorem 5.1 that set-level features add task-relevant information is not empirically supported.
  4. [Section 5 and Appendix A.2.3] The statement of Theorem 5.3 and its proof are not self-consistent about the dimension of theta. In Eq. (10), L_cr(theta^T Z_sup, Y_sup) and the surrounding text call theta a learnable matrix, whereas F_gamma = {Z -> theta^T Z : theta Sigma theta^T <= gamma} and the Rademacher proof treat theta as a row vector. The proof also applies Lemma A.1 directly to the linear class F_gamma without accounting for the Lipschitz constant of the cross-entropy loss, so the displayed bound on |R - hat_R| does not follow as written. The authors should either state the vector case with a scalar-output score and appropriate loss reduction, or adapt the Rademacher bound to the matrix-valued setting.
minor comments (7)
  1. [Algorithm 1, lines 5 and 9] Line 9 uses phi for the set-level projector, but Section 4.2 defines a separate projector psi for set representations; line 5 should be phi and line 9 should be psi.
  2. [Algorithm 1, input line] The input is written as G = {V, E, Z, A}, but the feature matrix is denoted X in Section 3 and Z is later defined as the final embedding, so the input notation should be corrected.
  3. [Section 4.3] The sentence saying that uniform alpha and beta 'ensure that Z_spt is evenly partitioned into NQ parts' is misleading; uniform marginals equate the total mass of each support point, not a partition of the support set.
  4. [Tables 3-5] Several entries are malformed or appear transposed: for example, the X-FNC row on ogbn-products shows '65.72 3.96' and '50.12 2.16' without the plus-minus sign, and some TLP values coincide with TENT values (e.g., TLP on WikiCS is 70.35). The tables need proofreading.
  5. [Reproducibility statement] The text says 'Our code can be found here' but does not provide an actual URL, so the reproducibility claim cannot be verified.
  6. [Section 6.2] The paper lists several supervised graph meta-learning baselines (G-Meta, TEG, COSMIC, etc.) alongside the unsupervised STAR; it should state explicitly whether these baselines use base-class labels during meta-training, since the unsupervised setting is one of the paper's main selling points.
  7. [Abstract] The phrase 'require necessitate' contains a duplicated wording and should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: STAR's empirical results are benchmark-anchored and not derived from its own definitions; the theoretical claims are vacuous and the OT calibration has a dimensional bug, but these are correctness issues, not circular reasoning.

full rationale

The paper's central empirical claim (state-of-the-art accuracy on eight datasets, Tables 3-5) is evaluated against external baselines and does not reuse fitted values or self-citations as evidence. The set-level feature construction (Section 4.2) is an independent design choice: top-k retrieval in an augmented view and set contrastive loss are stated as assumptions and validated indirectly (Fig. 4), not derived from the target results. The optimal-transport calibration (Section 4.3) is a standard Sinkhorn procedure; as written it is dimensionally inconsistent (\hat Z_spt = \lambda^\top Z_spt has NQ rows while support labels have length NK), which is a reproducibility/correctness defect, not a circular reduction. The theoretical section is largely vacuous: Theorem 5.1 is the data-processing inequality applied to the definitional concatenation Z = \tilde H || \tilde S, Corollary 5.2 contains a sign/notation error in the proof, and Theorem 5.3 is a generic Rademacher bound for linear classifiers. None of these constitute an input-output equivalence, fitted-parameter-as-prediction, or load-bearing self-citation. Self-citations (e.g., Meta-GPS) are used as baselines and context, not to justify STAR's claims. Therefore no significant circularity is present.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central empirical result relies on several hyperparameters that are chosen by hand or grid search and not fully reported. The theoretical results rely on standard inequalities plus two domain assumptions: that OT calibration is beneficial and that top-k neighbor sets are semantically coherent. No new entities are postulated.

free parameters (7)
  • top-k set size k = 20
    Chosen via sensitivity analysis in Fig. 3(b); controls how many neighbors form a set and affects set-level contrastive learning.
  • GCL temperature tau = 0.5
    Set for all contrastive losses; standard value, but not derived.
  • hidden dimension d' = 16
    Used for SGC, projectors, and set function; selected without reported search.
  • number of SGC layers ell = 2
    Encoder depth; fixed in implementation.
  • augmentation ratios (edge dropping, feature masking) = grid searched from 0 to 0.4
    Reported only as a range; exact values per dataset not given.
  • OT entropy regularization epsilon = not specified
    Controls smoothness of Sinkhorn plan; no value or search reported.
  • Sinkhorn iteration count = not specified
    Number of iterations to approximate lambda* is not reported.
assumptions (6)
  • standard math Data processing inequality and non-negativity of conditional mutual information
    Used in the proof of Theorem 5.1 (Appendix A.2.1) to show I(Z;T) >= I(~H;T) and >= I(~S;T).
  • standard math Hellman-Raviv inequality P(Z) <= H(Y|Z)/2
    Used in the proof of Corollary 5.2 (Eq. 17) to convert conditional entropy into a bound on Bayes error.
  • standard math Uniform deviation bound for Rademacher complexity (Lemma A.1 from Bartlett and Mendelson)
    Used in the proof of Theorem 5.3 to bound the generalization gap.
  • domain assumption Entropy-regularized optimal transport with uniform marginals produces a support-to-query calibration that helps classification
    Section 4.3 assumes the OT plan can move the support distribution onto the query distribution; the paper provides only a density visualization as evidence.
  • domain assumption Nearest-neighbor sets in augmented views are label-homogeneous (homophily in embedding space)
    Section 4.2 constructs positive set pairs from top-k dot-product neighbors across augmentations; the strategy relies on these neighbors sharing semantics with the target node.
  • ad hoc to paper Concatenation of instance and set embeddings is lossless for entropy
    Eq. 13 in Appendix A.2.1 sets H(Z) = H(~H, ~S) for Z = ~H||~S; true for deterministic concatenation but used to frame the mutual information comparison.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Unsupervised Graph Few-shot Learning via Set Functions and Optimal Transport." pith.science (2026). https://pith.science/paper/6PPOYUG6

@misc{pith2026250105635,
  author       = {Pith},
  title        = {Pith review of: Enhancing Unsupervised Graph Few-shot Learning via Set Functions and Optimal Transport},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6PPOYUG6}},
  note         = {Machine review of arXiv:2501.05635}
}
read the original abstract

Graph few-shot learning has garnered significant attention for its ability to rapidly adapt to downstream tasks with limited labeled data, sparking considerable interest among researchers. Recent advancements in graph few-shot learning models have exhibited superior performance across diverse applications. Despite their successes, several limitations still exist. First, existing models in the meta-training phase predominantly focus on instance-level features within tasks, neglecting crucial set-level features essential for distinguishing between different categories. Second, these models often utilize query sets directly on classifiers trained with support sets containing only a few labeled examples, overlooking potential distribution shifts between these sets and leading to suboptimal performance. Finally, previous models typically require necessitate abundant labeled data from base classes to extract transferable knowledge, which is typically infeasible in real-world scenarios. To address these issues, we propose a novel model named STAR, which leverages Set funcTions and optimAl tRansport for enhancing unsupervised graph few-shot learning. Specifically, STAR utilizes expressive set functions to obtain set-level features in an unsupervised manner and employs optimal transport principles to align the distributions of support and query sets, thereby mitigating distribution shift effects. Theoretical analysis demonstrates that STAR can capture more task-relevant information and enhance generalization capabilities. Empirically, extensive experiments across multiple datasets validate the effectiveness of STAR. Our code can be found here.

Figures

Figures reproduced from arXiv: 2501.05635 by the authors.

Figure 1
Figure 1. We observe a clear distribution shift between the support [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The overall framework of our model. where contrastive learning is applied, followed by 𝐿2 normalization, i.e., H′ =𝜙 (H), H′ =H′ /||H′ ||2. We consider pairs of nodes originating from the same node as positive samples, and pairs from different nodes as negative samples. Thus, the instance-level graph contrastive loss for two graphs with 2𝑛 nodes is defined as: L𝑖𝑛𝑠 = − 1 2𝑛 ∑︁2𝑛 𝑖=1 log exp( (H′ 𝑖 · H′ 𝑗 )/𝜏) Í 𝑘,𝑘≠… view at source ↗
Figure 3
Figure 3. (a): Distribution of support and query set after per [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Model performance varies with epochs across two [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 54 canonical work pages

  1. [1]

    Martin Arjovsky, Soumith Chintala, and Léon Bottou. 2017. Wasserstein genera- tive adversarial networks. In ICML

  2. [2]

    Peter L Bartlett and Shahar Mendelson. 2002. Rademacher and Gaussian complex- ities: Risk bounds and structural results. Journal of Machine Learning Research 3 (2002), 463–482

  3. [3]

    Aleksandar Bojchevski and Stephan Günnemann. 2018. Deep gaussian embedding of graphs: Unsupervised inductive learning via ranking. In ICLR

  4. [4]

    Liqun Chen, Yizhe Zhang, Ruiyi Zhang, Chenyang Tao, Zhe Gan, Haichao Zhang, Bai Li, Dinghan Shen, Changyou Chen, and Lawrence Carin. 2019. Improving sequence-to-sequence learning via optimal transport. In ICLR

  5. [5]

    Nicolas Courty, Rémi Flamary, Amaury Habrard, and Alain Rakotomamonjy. 2017. Joint distribution optimal transportation for domain adaptation. In NeurIPS

  6. [6]

    Nicolas Courty, Rémi Flamary, Devis Tuia, and Alain Rakotomamonjy. 2016. Optimal transport for domain adaptation. IEEE TPAMI 39, 9 (2016), 1853–1865

  7. [7]

    Marco Cuturi. 2013. Sinkhorn distances: Lightspeed computation of optimal transport. In NeurIPS

  8. [8]

    Kaize Ding, Jianling Wang, Jundong Li, Kai Shu, Chenghao Liu, and Huan Liu

Show all 64 references
  1. [9]

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. 2017. Model-agnostic meta- learning for fast adaptation of deep networks. In ICML

  2. [10]

    Jonathan Gordon, John Bronskill, Matthias Bauer, Sebastian Nowozin, and Richard E Turner. 2019. Meta-learning probabilistic inference for prediction. In ICLR

  3. [11]

    Aditya Grover and Jure Leskovec. 2016. node2vec: Scalable feature learning for networks. In SIGKDD

  4. [12]

    Dandan Guo, Long Tian, Minghe Zhang, Mingyuan Zhou, and Hongyuan Zha

  5. [13]

    Dandan Guo, Long Tian, He Zhao, Mingyuan Zhou, and Hongyuan Zha. 2022. Adaptive distribution calibration for few-shot learning with hierarchical optimal transport. In NeurIPS

  6. [14]

    Martin Hellman and Josef Raviv. 1970. Probability of error, equivocation, and the Chernoff bound. IEEE Transactions on Information Theory 16, 4 (1970), 368–372

  7. [15]

    Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. 2020. Open graph benchmark: Datasets for machine learning on graphs. In NeurIPS

  8. [16]

    Weihua Hu, Bowen Liu, Joseph Gomes, Marinka Zitnik, Percy Liang, Vijay Pande, and Jure Leskovec. 2020. Strategies for pre-training graph neural networks. In ICLR

  9. [17]

    Kexin Huang and Marinka Zitnik. 2020. Graph meta learning via local subgraphs. In NeurIPS

  10. [18]

    Maximilian Ilse, Jakub Tomczak, and Max Welling. 2018. Attention-based deep multiple instance learning. In ICML

  11. [19]

    Sungwon Kim, Junseok Lee, Namkyeong Lee, Wonjoong Kim, Seungyoon Choi, and Chanyoung Park. 2023. Task-Equivariant Graph Few-shot Learning. In SIGKDD

  12. [20]

    Diederik P Kingma and Jimmy Ba. 2015. Adam: A method for stochastic opti- mization. In ICLR

  13. [21]

    Thomas N Kipf and Max Welling. 2017. Semi-supervised classification with graph convolutional networks. In ICLR

  14. [22]

    Dong Bok Lee, Seanie Lee, Joonho Ko, Kenji Kawaguchi, Juho Lee, and Sung Ju Hwang. 2023. Self-Supervised Set Representation Learning for Unsupervised Meta-Learning. In ICLR

  15. [23]

    Juho Lee, Yoonho Lee, Jungtaek Kim, Adam Kosiorek, Seungjin Choi, and Yee Whye Teh. 2019. Set transformer: A framework for attention-based permutation-invariant neural networks. In ICML

  16. [24]

    Seanie Lee, Bruno Andreis, Kenji Kawaguchi, Juho Lee, and Sung Ju Hwang. 2022. Set-based meta-interpolation for few-task meta-learning. In NeurIPS

  17. [25]

    Mengyu Li, Yonghao Liu, Fausto Giunchiglia, Xiaoyue Feng, and Renchu Guan

  18. [26]

    Yonghao Liu, Fausto Giunchiglia, Lan Huang, Ximing Li, Xiaoyue Feng, and Renchu Guan. 2025. A Simple Graph Contrastive Learning Framework for Short Text Classification. In AAAI

  19. [27]

    Yonghao Liu, Renchu Guan, Fausto Giunchiglia, Yanchun Liang, and Xiaoyue Feng. 2021. Deep attention diffusion graph neural networks for text classification. In EMNLP

  20. [28]

    Yonghao Liu, Lan Huang, Bowen Cao, Ximing Li, Fausto Giunchiglia, Xiaoyue Feng, and Renchu Guan. 2024. A Simple but Effective Approach for Unsupervised Few-Shot Graph Classification. In WWW

  21. [29]

    Yonghao Liu, Lan Huang, Fausto Giunchiglia, Xiaoyue Feng, and Renchu Guan

  22. [30]

    Yonghao Liu, Mengyu Li, Ximing Li, Fausto Giunchiglia, Xiaoyue Feng, and Renchu Guan. 2022. Few-shot node classification on attributed networks with graph meta-learning. In SIGIR

  23. [31]

    Yonghao Liu, Mengyu Li, Ximing Li, Lan Huang, Fausto Giunchiglia, Yanchun Liang, Xiaoyue Feng, and Renchu Guan. 2024. Meta-GPS++: Enhancing Graph Meta-Learning with Contrastive Learning and Self-Training. ACM TKDD 18, 9 (2024), 1–30

  24. [32]

    Yonghao Liu, Mengyu Li, Di Liang, Ximing Li, Fausto Giunchiglia, Lan Huang, Xiaoyue Feng, and Renchu Guan. 2024. Resolving Word Vagueness with Scenario- guided Adapter for Natural Language Inference. In IJCAI

  25. [33]

    Improved Graph Contrastive Learning for Short Text Classification. In AAAI

  26. [34]

    Yonghao Liu, Di Liang, Fang Fang, Sirui Wang, Wei Wu, and Rui Jiang. 2023. Time-aware multiway adaptive fusion network for temporal knowledge graph question answering. In ICASSP. 1–5

  27. [35]

    Yonghao Liu, Di Liang, Mengyu Li, Fausto Giunchiglia, Ximing Li, Sirui Wang, Wei Wu, Lan Huang, Xiaoyue Feng, and Renchu Guan. 2023. Local and Global: Temporal Question Answering via Information Fusion. In IJCAI

  28. [36]

    Péter Mernyei and Cătălina Cangea. 2020. Wiki-cs: A wikipedia-based benchmark for graph neural networks. arXiv preprint arXiv:2007.02901 (2020)

  29. [37]

    Yonghao Liu, Mengyu Li, Wei Pang, Fausto Giunchiglia, Lan Huang, Xiaoyue Feng, and Renchu Guan. 2025. Boosting Short Text Classification with Multi- Source Information Exploration and Dual-Level Contrastive Learning. In AAAI

  30. [38]

    Stylianos Poulakakis-Daktylidis and Hadi Jamali-Rad. 2024. BECLR: Batch En- hanced Contrastive Few-Shot Learning. In ICLR

  31. [39]

    Tim Salimans, Han Zhang, Alec Radford, and Dimitris Metaxas. 2018. Improving GANs using optimal transport. In ICLR

  32. [40]

    Oleksandr Shchur, Maximilian Mumme, Aleksandar Bojchevski, and Stephan Günnemann. 2018. Pitfalls of graph neural network evaluation. arXiv preprint arXiv:1811.05868 (2018). KDD ’25, August 3–7, 2025, Toronto, ON, Canada Yonghao Liu et al

  33. [41]

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. 2014. Deepwalk: Online learning of social representations. In SIGKDD

  34. [42]

    Jake Snell, Kevin Swersky, and Richard Zemel. 2017. Prototypical networks for few-shot learning. In NeurIPS

  35. [43]

    Zhen Tan, Ruocheng Guo, Kaize Ding, and Huan Liu. 2023. Virtual node tuning for few-shot node classification. In SIGKDD. 2177–2188

  36. [44]

    Zhen Tan, Song Wang, Kaize Ding, Jundong Li, and Huan Liu. 2022. Transductive linear probing: a novel framework for few-shot node classification. In LoG

  37. [45]

    Hantao Shu, Jingtian Zhou, Qiuyu Lian, Han Li, Dan Zhao, Jianyang Zeng, and Jianzhu Ma. 2021. Modeling gene regulatory networks using neural network architectures. Nature Computational Science 1, 7 (2021), 491–501

  38. [46]

    Eleni Triantafillou, Tyler Zhu, Vincent Dumoulin, Pascal Lamblin, Utku Evci, Kelvin Xu, Ross Goroshin, Carles Gelada, Kevin Swersky, Pierre-Antoine Man- zagol, et al. 2020. Meta-dataset: A dataset of datasets for learning to learn from few examples. In ICLR

  39. [47]

    Mikaela Angelina Uy, Quang-Hieu Pham, Binh-Son Hua, Thanh Nguyen, and Sai-Kit Yeung. 2019. Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data. In ICCV

  40. [48]

    Oriol Vinyals, Samy Bengio, and Manjunath Kudlur. 2016. Order matters: Se- quence to sequence for sets. In ICLR

  41. [49]

    Luis Caicedo Torres, Luiz Manella Pereira, and M Hadi Amini. 2021. A survey on optimal transport for machine learning: Theory and applications. arXiv preprint arXiv:2106.01963 (2021)

  42. [50]

    Song Wang, Yushun Dong, Kaize Ding, Chen Chen, and Jundong Li. 2023. Few- shot node classification with extremely weak supervision. In WSDM

  43. [51]

    Song Wang, Zhen Tan, Huan Liu, and Jundong Li. 2023. Contrastive Meta- Learning for Few-shot Node Classification. In SIGKDD. 2386–2397

  44. [52]

    Felix Wu, Amauri Souza, Tianyi Zhang, Christopher Fifty, Tao Yu, and Kilian Weinberger. 2019. Simplifying graph convolutional networks. In ICML

  45. [53]

    Song Wang, Kaize Ding, Chuxu Zhang, Chen Chen, and Jundong Li. 2022. Task- adaptive few-shot node classification. In SIGKDD

  46. [54]

    Huaxiu Yao, Long-Kai Huang, Linjun Zhang, Ying Wei, Li Tian, James Zou, Junzhou Huang, et al. 2021. Improving generalization in meta-learning via task augmentation. In ICML

  47. [55]

    Huaxiu Yao, Linjun Zhang, and Chelsea Finn. 2022. Meta-learning with fewer tasks through task interpolation. In ICLR

  48. [56]

    Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. 2020. Graph contrastive learning with augmentations. In NeurIPS

  49. [57]

    Zhilin Yang, William Cohen, and Ruslan Salakhudinov. 2016. Revisiting semi- supervised learning with graph embeddings. In ICML

  50. [58]

    Manzil Zaheer, Satwik Kottur, Siamak Ravanbakhsh, Barnabas Poczos, Russ R Salakhutdinov, and Alexander J Smola. 2017. Deep sets. In NeurIPS

  51. [59]

    Chuxu Zhang, Kaize Ding, Jundong Li, Xiangliang Zhang, Yanfang Ye, Nitesh V Chawla, and Huan Liu. 2022. Few-shot learning on graphs. arXiv preprint arXiv:2203.09308 (2022)

  52. [60]

    Fan Zhou, Chengtai Cao, Kunpeng Zhang, Goce Trajcevski, Ting Zhong, and Ji Geng. 2019. Meta-gnn: On few-shot node classification in graph meta-learning. In CIKM. A SUPPLEMENTARY MATERIAL A.1 Complexity Analysis In this section, we conduct a detailed analysis of the time comple...

  53. [61]

    Mikhail Yurochkin, Sebastian Claici, Edward Chien, Farzaneh Mirzazadeh, and Justin M Solomon. 2019. Hierarchical optimal transport for document represen- tation. In NeurIPS

  54. [2020]

    Graph prototypical networks for few-shot learning on attributed networks. In CIKM

  55. [2022]

    Learning prototype-oriented set representations for meta-learning. In ICLR

  56. [2024]

    arxiv preprint arXiv:2405.11524 (2024)

    Simple-Sampling and Hard-Mixup with Prototypes to Rebalance Con- trastive Learning for Text Classification. arxiv preprint arXiv:2405.11524 (2024)

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.