Pith. sign in

REVIEW 2 major objections 7 minor 54 references

Instance-Aware Graph Prompt Learning

T0 review · 2 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read The paper claims that generating a distinct prompt for every input graph, via a parameter-efficient PHM bottleneck and vector quantization, outperforms fixed task-level prompts on few-shot graph classification while keeping the pretrained…

desk verdict A solid few-shot graph prompting extension with a thorough experimental appendix, but a dimension mismatch in the central injection step and an overstated novelty claim need fixing before it is citable as written. read the letter →

arxiv 2411.17676 v1 pith:U7MGWWM5 submitted 2024-11-26 cs.LG

classification cs.LG
keywords graphpromptlearninginstance-awarepromptsneuralnetworksvectorquantizationfew-shotparameter-efficientfine-tuningmolecularpropertypredictionout-of-distributiongeneralization
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

Existing graph prompt learning methods attach one fixed prompt to every graph in a task, even though graphs within a task differ widely; the paper argues this static prompt is the key limitation and that prompts should be generated per input instance. IA-GPL therefore produces a distinct prompt for each node by passing the frozen GNN's hidden representations through a parameter-efficient bottleneck built from parameterized hypercomplex multiplication layers, then quantizes those prompt vectors through a small trainable codebook updated by exponential moving average. The claimed result is that this frozen-backbone method beats fixed-prompt baselines (GPF, GPF-plus, All-in-One) and often approaches or exceeds full fine-tuning on eight molecular datasets and three citation networks under few-shot and full-shot settings, with the largest gains in out-of-domain scaffold splits. If the claim holds, prompt tuning on graphs becomes per-instance adaptation rather than task-level patching, which would matter for any application where a single prompt must cover structurally heterogeneous inputs.

What carries the argument

The load-bearing object is the prompt generator $g_\Phi(H) = \mathrm{UpProject}(\mathrm{ReLU}(\mathrm{DownProject}(H)))$, with both projectors implemented as PHM layers. A PHM layer replaces the full weight matrix $W$ with a sum of Kronecker products $M = \sum_{i=1}^n A_i \otimes S_i$, so the number of trainable parameters scales roughly as $1/n$ of a fully connected layer. The intermediate per-node prompts $P_c = g_\Phi(H)$ are then mapped through vector quantization: for each prompt, $M$ codebook vectors are sampled from a multinomial distribution over negative squared Euclidean distances scaled by a temperature, and the quantized prompt $p_q$ is the average of those sampled vectors. The final prompt is $p_f = p_q + \beta p_s$, where $p_s$ is a shared learnable prompt. Codebook vectors are updated by exponential moving average rather than backpropagation, which the paper argues prevents representation collapse and stabilizes training; the training objective adds a consistency loss $\lambda \sum_i \|p_{q,i} - p_{c,i}\|_2^2$ to the cross-entropy loss.

What would settle it

Run the released implementation on a molecular dataset with 9-dimensional raw features and a hidden dimension $d>9$: if the addition $X_p=X+P$ in Eq. (7) fails or requires an undocumented projection, then the published mechanism is not reproducible as written. Independently, replace the vector-quantization step with simple averaging of the PHM outputs while keeping all other hyperparameters fixed; if out-of-domain ROC-AUC on scaffold splits does not drop, the paper's central attribution of OOD gains to the codebook is falsified.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that instance-aware prompts, generated from the input graph's own hidden representation rather than learned as a single task-level vector, improve graph prompt learning. The authors report that IA-GPL achieves the highest average ROC-AUC among graph prompting methods on eight MoleculeNet benchmarks under both random split (68.46 in 50-shot) and scaffold split (61.59 in 50-shot), and under full-shot scaffold split (74.93), along with the best accuracy on Cora, CiteSeer, and PubMed in the 100-shot node-level setting. They further report that IA-GPL obtains the best result in 27 of 32 configurations tested across five pretraining strategies. They attribute the improvement to two design choices acting together: the PHM bottleneck keeps per-instance generation cheap, and vector quantization with EMA-updated codebooks forces distinct, stable prompts whose clustering transfers better to out-of-distribution instances.

Load-bearing premise

The load-bearing premise is that a prompt vector computed in the GNN's hidden space can be added directly to the raw input features ($X_p = X+P$) and the frozen backbone will still read the result, yet on the molecular datasets raw features are 9-dimensional while hidden prompts are $d$-dimensional, so the stated injection is not dimensionally well-defined.

Editorial extensions

If this is right

  • IA-GPL reports the best average ROC-AUC among graph prompting methods on eight molecular datasets in 50-shot random split (68.46) and scaffold split (61.59), and in full-shot scaffold split (74.93).
  • On 100-shot node-level tasks over Cora, CiteSeer, and PubMed, IA-GPL reports the highest accuracy among prompting baselines, with an average of 72.06.
  • Across five pretraining objectives, IA-GPL is reported as the best prompting method in 27 of 32 configurations, so the gains are not tied to one particular pretrained backbone.
  • With roughly 20K trainable parameters (about 1.08% of fine-tuning's 1.86M), the method keeps the GNN frozen and uses comparable training time and slightly less GPU memory than fine-tuning.
  • The performance gap over static prompts is larger under scaffold split than under random split, which the paper takes as evidence that vector quantization's clustering property helps out-of-domain generalization.

Reading between the lines

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

  • The same PHM-bottleneck-plus-codebook recipe could be dropped into other frozen encoders, such as text or vision transformers where fixed soft prompts are standard; the paper's mechanism does not depend on graph-specific message passing beyond the hidden representations it starts from.
  • The paper leaves the dimensionality of the injection $X_p = X+P$ unspecified: molecular node features are 9-dimensional while hidden prompts live in the GNN hidden dimension $d$. A concrete implementation would need an explicit projection or padding, and the stated equation is not self-consistent as written.
  • One can test directly whether the codebook vectors carry semantic meaning by probing whether clusters of quantized prompts align with chemical functional groups across molecules; the paper's t-SNE suggests clustering but does not quantify this.
  • An ablation fixing codebook size and varying only the sampling temperature would separate the contribution of stochastic sampling from the contribution of quantization itself.
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

2 major / 7 minor

Summary. IA-GPL is a graph prompting method that replaces the static task-level prompts of prior work (GPF, GPF-plus, GPPT, All-in-One) with node-level prompts generated per input instance. The pipeline is: (i) a frozen pre-trained GNN encodes the input graph to node representations H; (ii) a parameter-efficient bottleneck of PHM layers maps H to intermediate prompts p_c; (iii) vector quantization over a learnable codebook with multinomial sampling and an EMA update yields quantized prompts p_q, combined with a shared static prompt p_s; (iv) the final node-level prompt P is added to the original node features (Eq. 7), and the prompted graph is fed into the frozen backbone for classification. The paper reports full-shot and few-shot experiments on eight MoleculeNet datasets under random and scaffold splits, three citation datasets, and PPI, with five pre-training strategies, comparing against supervised, fine-tuning, linear-probing, and prompt-learning baselines, plus ablations, efficiency analysis, codebook visualization, and hyperparameter-sensitivity studies.

Significance. The contribution is meaningful and the experimental effort is unusually extensive: 12 datasets, 5 pre-training strategies, two split regimes, few-shot and full-shot regimes, plus ablations and efficiency measurements. If the method description is made precise, the empirical claim is plausible: average gains over the strongest prompting baselines are consistent (Table 2: 61.59 vs. 60.09 for GPF under scaffold 50-shot; Table 5: 74.93 vs. 73.91 for GPF-plus under full-shot scaffold), the few-shot gains are in the regime where prompt methods are claimed to help, and the VQ clustering rationale is a concrete, falsifiable mechanism for the OOD gains. The method is genuinely parameter-efficient (Table 3: ~20K tunable parameters vs. 1.86M for fine-tuning) and the ablation isolates the PHM and VQ components. The main caveat is that the paper's own description is internally inconsistent at the central injection step and in the backbone specification, so the reported numbers cannot currently be reproduced or even unambiguously interpreted; this is fixable but must precede any final judgment.

major comments (2)
  1. [Section 4.1, Eq. (7)] Section 4.1, Eq. (7): the central prompt-injection step is not well-defined as written. The text defines H in R^{|V| x d} as the output of the frozen GNN, defines P = g_Phi(H), and then says P in R^{|V| x d} is 'matching the same shape as X so that they can be added back to the original node features,' with X_p = X + P. The same symbol d is used in Section 3 for the raw node-feature dimension, which is 9 for every MoleculeNet dataset in Table 4, while the frozen 5-layer GIN backbone described in Appendix C operates in a far larger hidden dimension. Therefore P cannot simultaneously live in the hidden dimension of H and have the same shape as X. The authors must either (a) explicitly project the prompt to the input feature dimension d_X, in which case the statements 'projected back to d dimensions,' the bottleneck asymmetry, the codebook dimension in Eqs. (13)-(15), and the consistency loss in Eq. (17) all need to be restated in the prompt dimension, or (b) state that prompts are injected into a hidden representation, which contradicts the text's repeated claim that prompts are added to the original node features and would change the basis of the comparison to input-space methods such as GPF. Because the prompted graph G_p = (X_p, A) is the input to the frozen backbone in every experiment, this ambiguity is load-bearing for the method description and must be resolved (including with code or a pseudocode block) before the contribution can be evaluated.
  2. [Section 5.1 vs. Appendix C] Section 5.1 vs. Appendix C: the experimental backbone is described inconsistently. Section 5.1 states 'GCN is adopted as our backbone model,' while Appendix C states that for molecular datasets 'we adopt the widely used 5-layer GIN' and for citation networks 'we adopt 2-layer Graph Transformers'; Section 6 and the ablation discussion in Section 5.3 also refer to 'the huge pre-trained GIN model.' The pre-trained checkpoints and frozen backbone determine H, the space in which prompts are generated, as well as the validity of the comparison across baselines. A single consistent statement of which architecture(s) generated Tables 1, 2, 5-9 is required for reproducibility, and the authors should also clarify which pre-trained checkpoints (e.g., from Hu et al. 2020) were used.
minor comments (7)
  1. [Section 5.2, Tables 1-2] The text in Section 5.2 states that IA-GPL 'attains optimal results across these eight datasets' under the scaffold 50-shot setting, but in Table 2 GPF achieves a higher ROC-AUC on ClinTox (65.18 +/- 1.76 vs. 63.28 +/- 3.52); similarly, in Table 1, GPF-plus exceeds IA-GPL on MUV (59.93 +/- 0.83 vs. 59.32 +/- 1.13). Suggest rephrasing these claims to 'best average' or explicitly reporting per-dataset exceptions.
  2. [Section 6 and Appendix C] The code repository is referenced twice ('You can also visit our code repository'), but no URL, repository name, or commit hash is provided; given the ambiguity in Eq. (7), providing the actual implementation of the injection step would materially resolve the main concern.
  3. [Section 4.3, Eq. (14)] Eq. (14) draws from a 'Multinomial distribution over the logits,' but a Multinomial distribution requires probabilities; the authors should state explicitly that the negative-distance scores are normalized (e.g., by a softmax over -d_i^c/tau) before sampling, and clarify the role of tau in that normalization.
  4. [Section 4.4, Eqs. (18)-(19)] In the EMA update, the count c_j updated in Eq. (18) is used as the denominator in Eq. (19) for the same batch; please clarify whether the count used in the mean is the updated running count or the per-batch count, since this affects the effective codebook learning dynamics.
  5. [Section 5.3 and Table 3] The inference procedure requires two forward passes through the frozen GNN, one on G to produce H for prompt generation and one on G_p for prediction, but Table 3 reports only training time and memory; a statement of this inference overhead (and whether it is included in the reported training time) would improve the efficiency analysis.
  6. [Throughout] Several typos and inconsistencies should be corrected: 'IA-GPF' for 'IA-GPL' in Section 5.2; 'instance-agonist' for 'instance-agnostic' in Section 4.3 and Figure 3; 'virgina.edu' in the author block; and the first author's email domain (brandeis.edu) differs from the stated affiliation (University of Connecticut).
  7. [Section 5.2, Tables 1-2] No statistical significance tests are reported for the few-shot tables, where several comparisons fall within one standard deviation (e.g., Table 2, HIV: 66.52 +/- 2.10 vs. 65.59 +/- 2.31); paired significance tests across the five runs would strengthen the superiority claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: IA-GPL's instance-aware prompt generation is an empirical architecture whose parameters are trained on labeled downstream data and evaluated on held-out external benchmarks.

full rationale

I walked the derivation chain from the frozen GNN encoding in Eq. (6) through the prompt generation, quantization, and optimization in Eqs. (7)-(19). The instance-aware prompt P is produced by a PHM bottleneck gΦ applied to the frozen GNN's node representations H, then quantized through codebook vectors and combined with a static prompt. None of these quantities is defined in terms of the target labels or the test predictions; they are functions of the input graph and trainable parameters. The loss in Eq. (17) combines cross-entropy with a consistency regularizer between quantized and intermediate prompts, and the EMA updates in Eqs. (18)-(19) update codebooks from batch statistics. This is standard supervised training, not a fitted parameter being renamed as a prediction. Evaluation is conducted against external baselines on held-out test splits under random and scaffold splits and multiple pretraining strategies, so the central claim is empirically grounded rather than circular. The paper relies on external prior work for PHM layers, vector quantization, EMA, and graph prompting baselines, and I found no load-bearing self-citation chain or imported uniqueness claim. The dimension ambiguity around Eq. (7), where H and X are both denoted R^{|V|×d}, is a correctness or implementation concern about whether the prompt injection is well-defined as written, but it is not a circularity: resolving the projection does not make the method's output equal to its input by definition. Therefore the appropriate score is 0.

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

The central claim depends on several hand-set hyperparameters, the validity of the PHM and VQ machinery from prior work, and the unstated assumption that prompted features remain valid inputs to the frozen GNN. Most components are adapted from cited work rather than derived in this paper.

free parameters (7)
  • Codebook size K = 10, 20, or 50 depending on dataset (Table 10)
    The number of codebook vectors is chosen per dataset and split by grid search; it controls prompt diversity and capacity.
  • Number of multinomial samples M = 5, 10, or 50 per dataset (Table 10)
    Number of codebook vectors averaged per prompt; sampled with replacement from the codebook.
  • Temperature tau = not specified
    Temperature in Eq. (13) controls sampling sharpness, but no default or tuned value is reported.
  • Balancing hyperparameter beta = not specified
    Weight for the instance-agnostic prompt ps in Eq. (16); no value or tuning range is given.
  • Consistency loss weight lambda = 0.01
    Weight for the ||pq - pc||^2 term in Eq. (17); set to 0.01 but sensitivity is not analyzed.
  • EMA decay alpha = 0.99
    Moving average coefficient in Eqs. (18) and (19); set to 0.99.
  • Bottleneck dimension d' = not specified
    Hidden width of the PHM down-and-up projectors; controls parameter count, but the exact value is not reported.
assumptions (6)
  • standard math PHM layer parameterization (Eqs. 10-12) is a valid low-rank replacement for a fully connected layer.
    Adopted from Zhang et al. (2020); the paper relies on this without proof.
  • domain assumption Vector quantization with EMA training provides stable, non-collapsed prompt prototypes.
    Used as a design assumption in Sections 4.3 and 4.4; cluster stability is asserted, not proven.
  • domain assumption Adding per-node prompts to the input features leaves the frozen GNN's feature distribution usable.
    The method rests on the premise that X + P in Eq. (7) remains a valid input to the frozen backbone; the paper does not justify this distributionally.
  • domain assumption Unifying node- and graph-level tasks into a general graph-level task via local subgraphs preserves task information.
    The evaluation protocol relies on this conversion, following Sun et al. (2023) and Liu et al. (2024), especially for node-level citation datasets.
  • domain assumption The pretrained GNN representations H are sufficiently informative to generate useful prompts.
    The prompt generator uses only H in Eqs. (6) to (8); if H lacks task-relevant information, prompts cannot recover it.
  • ad hoc to paper Xp = X + P is a well-defined operation (Eq. 7).
    The paper writes P in the hidden dimension d and X in the raw feature dimension, which for MoleculeNet datasets do not match; the paper implicitly assumes some unstated projection or alternative injection.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Instance-Aware Graph Prompt Learning." pith.science (2026). https://pith.science/paper/U7MGWWM5

@misc{pith2026241117676,
  author       = {Pith},
  title        = {Pith review of: Instance-Aware Graph Prompt Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U7MGWWM5}},
  note         = {Machine review of arXiv:2411.17676}
}
read the original abstract

Graph neural networks stand as the predominant technique for graph representation learning owing to their strong expressive power, yet the performance highly depends on the availability of high-quality labels in an end-to-end manner. Thus the pretraining and fine-tuning paradigm has been proposed to mitigate the label cost issue. Subsequently, the gap between the pretext tasks and downstream tasks has spurred the development of graph prompt learning which inserts a set of graph prompts into the original graph data with minimal parameters while preserving competitive performance. However, the current exploratory works are still limited since they all concentrate on learning fixed task-specific prompts which may not generalize well across the diverse instances that the task comprises. To tackle this challenge, we introduce Instance-Aware Graph Prompt Learning (IA-GPL) in this paper, aiming to generate distinct prompts tailored to different input instances. The process involves generating intermediate prompts for each instance using a lightweight architecture, quantizing these prompts through trainable codebook vectors, and employing the exponential moving average technique to ensure stable training. Extensive experiments conducted on multiple datasets and settings showcase the superior performance of IA-GPL compared to state-of-the-art baselines.

Figures

Figures reproduced from arXiv: 2411.17676 by the authors.

Figure 1
Figure 1. Two example molecules from the BBBP dataset. Molecule (a) with simple structures suffices with a universal prompt. However, molecule (b) with diverse atoms and intricate structures requires the use of instance-aware prompts. This approach not only preserves performance but also contributes to a reduction in training resource consumption. In the graph domain, prompt learning has recently demonstrated its potential as… view at source ↗
Figure 2
Figure 2. Comparison between different paradigms of graph representation learning. frozen GNN, which contain rich neighbor-aware information, further aiding in the prompt generation process. Thus in this work, we introduce IA-GPL, a novel methodology designed to address the aforementioned issue by generating prompts that leverage the distinctive features in individual instances. 3 Preliminaries Graphs. Let G = (V, E, X, A) re… view at source ↗
Figure 3
Figure 3. Overall Framework of IA-GPL. where the weight matrix W ∈ R k×d and the bias vector b ∈ R k are trainable parameters. We can control the number of parameters by controlling the hidden dimension d ′ , but it is a trade-off between performance and efficiency. In other words, it contradicts the original objective of prompt learning, which aims to reduce the number of trainable parameters, if we set d ′ large to maintain… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Codebook visualization. outperforms LP, whereas the reverse is observed for scaffold split. This observation confirms the negative transfer drawback associated with the "pre-training and fine-tuning" paradigm: the gap between pretext tasks and downstream tasks leads to…
Figure 5
Figure 5. Figure 5: Ablation study. Two characteristics of the learned codebooks are observed: (1) Samples corresponding to different atoms manifest substantial distinctions (i.e., the regions of samples in the plot). However, sam￾ples corresponding to the same atoms tend to ex￾hibit in p…
Figure 6
Figure 6. Figure 6: Impact of shot numbers. In terms of training efficiency, we compute the training time per epoch and GPU mem￾ory consumption on the ToxCast dataset us￾ing a single Nvidia RTX 3090. We keep all hyper-parameters the same including batch size, dimensions, etc. All-in-One i…
Figure 7
Figure 7. Figure 7: Impact of VQ hyperparameters. Impacts of the shot number. We study the impact of the number of shots on the BBBP and BACE datasets in the few-shot random split setting. We vary the number of shots within the range of [5,10,20,30] and results are illustrated in [PITH_F…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 41 canonical work pages

  1. [1]

    Extractive opinion summarization in quantized transformer spaces

    Stefanos Angelidis, Reinald Kim Amplayo, Yoshihiko Suhara, Xiaolan Wang, and Mirella Lapata. Extractive opinion summarization in quantized transformer spaces. Transactions of the Association for Computational Linguistics, 2021

  2. [2]

    Beit: Bert pre-training of image transformers

    Hangbo Bao, Li Dong, Songhao Piao, and Furu Wei. Beit: Bert pre-training of image transformers. In ICLR, 2021

  3. [3]

    Vector-quantized input-contextualized soft prompts for natural language understanding

    Rishabh Bhardwaj, Amrita Saha, Steven CH Hoi, and Soujanya Poria. Vector-quantized input-contextualized soft prompts for natural language understanding. In EMNLP, 2022

  4. [4]

    Enhancing graph neural network-based fraud detectors against camouflaged fraudsters

    Yingtong Dou, Zhiwei Liu, Li Sun, Yutong Deng, Hao Peng, and Philip S Yu. Enhancing graph neural network-based fraud detectors against camouflaged fraudsters. In CIKM, 2020

  5. [5]

    Multiscale vision transformers

    Haoqi Fan, Bo Xiong, Karttikeya Mangalam, Yanghao Li, Zhicheng Yan, Jitendra Malik, and Christoph Feichtenhofer. Multiscale vision transformers. In CVPR, 2021

  6. [6]

    Graph neural networks for social recommendation

    Wenqi Fan, Yao Ma, Qing Li, Yuan He, Eric Zhao, Jiliang Tang, and Dawei Yin. Graph neural networks for social recommendation. In WWW, 2019

  7. [7]

    Universal prompt tuning for graph neural networks

    Taoran Fang, Yunchao Zhang, Yang Yang, Chunping Wang, and Lei Chen. Universal prompt tuning for graph neural networks. In NeurIPS, 2023

  8. [8]

    Making pre-trained language models better few-shot learners

    Tianyu Gao, Adam Fisch, and Danqi Chen. Making pre-trained language models better few-shot learners. arXiv preprint arXiv:2012.15723, 2020

Show all 54 references
  1. [9]

    Bellis, A

    Anna Gaulton, Louisa J. Bellis, A. Patricia Bento, Jon Chambers, Mark Davies, Anne Hersey, Yvonne Light, Shaun McGlinchey, David Michalovich, Bissan Al-Lazikani, and John P. Overington. Chembl: A large-scale bioactivity database for drug discovery. Nucleic acids research, 2011

  2. [10]

    Quantization

    RM GRAY. Quantization. IEEE transactions on information theory, 1998

  3. [11]

    node2vec: Scalable feature learning for networks

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

  4. [12]

    Ppt: Pre-trained prompt tuning for few-shot learning

    Yuxian Gu, Xu Han, Zhiyuan Liu, and Minlie Huang. Ppt: Pre-trained prompt tuning for few-shot learning. arXiv preprint arXiv:2109.04332, 2021

  5. [13]

    Few-shot graph learning for molecular property prediction

    Zhichun Guo, Chuxu Zhang, Wenhao Yu, John Herr, Olaf Wiest, Meng Jiang, and Nitesh V Chawla. Few-shot graph learning for molecular property prediction. In WWW, 2021

  6. [14]

    A deep graph neural network-based mechanism for social recommendations

    Zhiwei Guo and Heng Wang. A deep graph neural network-based mechanism for social recommendations. IEEE Transactions on Industrial Informatics, 2020

  7. [15]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. 2017

  8. [16]

    Strategies for pre-training graph neural networks

    W Hu, B Liu, J Gomes, M Zitnik, P Liang, V Pande, and J Leskovec. Strategies for pre-training graph neural networks. In ICLR, 2020 a

  9. [17]

    Strategies for pre-training graph neural networks

    W Hu, B Liu, J Gomes, M Zitnik, P Liang, V Pande, and J Leskovec. Strategies for pre-training graph neural networks. In ICLR, 2020 b

  10. [18]

    Open graph benchmark: Datasets for machine learning on graphs

    Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. Open graph benchmark: Datasets for machine learning on graphs. arXiv preprint arXiv:2005.00687, 2020 c

  11. [19]

    Self-supervised learning on graphs: Deep insights and new direction

    Wei Jin, Tyler Derr, Haochen Liu, Yiqi Wang, Suhang Wang, Zitao Liu, and Jiliang Tang. Self-supervised learning on graphs: Deep insights and new direction. arXiv preprint arXiv:2006.10141, 2020

  12. [20]

    A comprehensive survey on deep graph representation learning

    Wei Ju, Zheng Fang, Yiyang Gu, Zequn Liu, Qingqing Long, Ziyue Qiao, Yifang Qin, Jianhao Shen, Fang Sun, Zhiping Xiao, et al. A comprehensive survey on deep graph representation learning. arXiv preprint arXiv:2304.05055, 2023

  13. [21]

    Semi-supervised classification with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016 a

  14. [22]

    Variational graph auto-encoders

    Thomas N Kipf and Max Welling. Variational graph auto-encoders. arXiv preprint arXiv:1611.07308, 2016 b

  15. [23]

    Prefix-tuning: Optimizing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. In ACL, 2021 a

  16. [24]

    Prefix-tuning: Optimizing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimizing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021 b

  17. [25]

    One for all: Towards training one graph model for all classification tasks

    Hao Liu, Jiarui Feng, Lecheng Kong, Ningyue Liang, Dacheng Tao, Yixin Chen, and Muhan Zhang. One for all: Towards training one graph model for all classification tasks. In ICLR, 2024

  18. [26]

    Indigo: Gnn-based inductive knowledge graph completion using pair-wise encoding

    Shuwen Liu, Bernardo Grau, Ian Horrocks, and Egor Kostylev. Indigo: Gnn-based inductive knowledge graph completion using pair-wise encoding. In NuerIPS, 2021 a

  19. [27]

    P -tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks

    Xiao Liu, Kaixuan Ji, Yicheng Fu, Weng Tam, Zhengxiao Du, Zhilin Yang, and Jie Tang. P -tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks. In ACL, 2022 a

  20. [28]

    P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks

    Xiao Liu, Kaixuan Ji, Yicheng Fu, Weng Tam, Zhengxiao Du, Zhilin Yang, and Jie Tang. P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks. In ACL, 2022 b

  21. [29]

    Pick and choose: A gnn-based imbalanced learning approach for fraud detection

    Yang Liu, Xiang Ao, Zidi Qin, Jianfeng Chi, Jinghua Feng, Hao Yang, and Qing He. Pick and choose: A gnn-based imbalanced learning approach for fraud detection. In WWW, 2021 b

  22. [30]

    Content matters: a gnn-based model combined with text semantics for social network cascade prediction

    Yujia Liu, Kang Zeng, Haiyang Wang, Xin Song, and Bin Zhou. Content matters: a gnn-based model combined with text semantics for social network cascade prediction. In PAKDD, 2021 c

  23. [31]

    Graphprompt: Unifying pre-training and downstream tasks for graph neural networks

    Zemin Liu, Xingtong Yu, Yuan Fang, and Xinming Zhang. Graphprompt: Unifying pre-training and downstream tasks for graph neural networks. In WWW, 2023

  24. [32]

    Large-scale comparison of machine learning methods for drug target prediction on chembl

    Andreas Mayr, Günter Klambauer, Thomas Unterthiner, Marvin Steijaert, Joerg Wegner, Hugo Ceulemans, Djork-Arné Clevert, and Sepp Hochreiter. Large-scale comparison of machine learning methods for drug target prediction on chembl. Chemical Science, 2018

  25. [33]

    o f, G \

    Roc \' o Mercado, Tobias Rastemo, Edvard Lindel \"o f, G \"u nter Klambauer, Ola Engkvist, Hongming Chen, and Esben Jannik Bjerrum. Graph networks for molecular design. Machine Learning: Science and Technology, 2021

  26. [34]

    Theory and experiments on vector quantized autoencoders

    Aurko Roy, Ashish Vaswani, Arvind Neelakantan, and Niki Parmar. Theory and experiments on vector quantized autoencoders. arXiv preprint arXiv:1805.11063, 2018

  27. [35]

    Exploiting cloze questions for few shot text classification and natural language inference

    Timo Schick and Hinrich Sch \"u tze. Exploiting cloze questions for few shot text classification and natural language inference. arXiv preprint arXiv:2001.07676, 2020

  28. [36]

    Teague Sterling and John J. Irwin. Zinc 15 – ligand discovery for everyone. Journal of Chemical Information and Modeling, 2015

  29. [37]

    Gppt: Graph pre-training and prompt tuning to generalize graph neural networks

    Mingchen Sun, Kaixiong Zhou, Xin He, Ying Wang, and Xin Wang. Gppt: Graph pre-training and prompt tuning to generalize graph neural networks. In KDD, 2022

  30. [38]

    All in one: Multi-task prompting for graph neural networks

    Xiangguo Sun, Hong Cheng, Jia Li, Bo Liu, and Jihong Guan. All in one: Multi-task prompting for graph neural networks. In KDD, 2023

  31. [39]

    Yijun Tian, Chuxu Zhang, Zhichun Guo, Chao Huang, Ronald Metoyer, and Nitesh V. Chawla. Reciperec: A heterogeneous graph learning model for recipe recommendation. In IJCAI, 2022

  32. [40]

    Visualizing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of machine learning research, 2008

  33. [41]

    Graph attention networks

    Petar Veli c kovi \'c , Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks. arXiv preprint arXiv:1710.10903, 2017

  34. [42]

    Deep graph infomax

    Petar Veli c kovi \'c , William Fedus, William L Hamilton, Pietro Li \`o , Yoshua Bengio, and R Devon Hjelm. Deep graph infomax. In ICLR, 2018

  35. [43]

    Moleculenet: a benchmark for molecular machine learning

    Zhenqin Wu, Bharath Ramsundar, Evan N Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S Pappu, Karl Leswing, and Vijay Pande. Moleculenet: a benchmark for molecular machine learning. Chemical science, 2018

  36. [44]

    Simgrace: A simple framework for graph contrastive learning without data augmentation

    Jun Xia, Lirong Wu, Jintao Chen, Bozhen Hu, and Stan Z Li. Simgrace: A simple framework for graph contrastive learning without data augmentation. In WWW, 2022

  37. [45]

    How powerful are graph neural networks? In ICLR, 2018

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks? In ICLR, 2018

  38. [46]

    Revisiting semi-supervised learning with graph embeddings

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

  39. [47]

    A comprehensive survey of graph neural networks for knowledge graphs

    Zi Ye, Yogan Jaya Kumar, Goh Ong Sing, Fengyan Song, and Junsong Wang. A comprehensive survey of graph neural networks for knowledge graphs. IEEE Access, 2022

  40. [48]

    Graph contrastive learning with augmentations

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

  41. [49]

    Graph contrastive learning with augmentations

    Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, and Yang Shen. Graph contrastive learning with augmentations. NeurIPS, 33: 0 5812--5823, 2020 b

  42. [50]

    Graph transformer networks

    Seongjun Yun, Minbyul Jeong, Raehyun Kim, Jaewoo Kang, and Hyunwoo J Kim. Graph transformer networks. 2019 a

  43. [51]

    Graph transformer networks

    Seongjun Yun, Minbyul Jeong, Raehyun Kim, Jaewoo Kang, and Hyunwoo J Kim. Graph transformer networks. NeurIPS, 2019 b

  44. [52]

    Beyond fully-connected layers with quaternions: Parameterization of hypercomplex multiplications with 1/n parameters

    Aston Zhang, Yi Tay, SHUAI Zhang, Alvin Chan, Anh Tuan Luu, Siu Hui, and Jie Fu. Beyond fully-connected layers with quaternions: Parameterization of hypercomplex multiplications with 1/n parameters. In ICLR, 2020

  45. [53]

    Graph contrastive learning with adaptive augmentation

    Yanqiao Zhu, Yichen Xu, Feng Yu, Qiang Liu, Shu Wu, and Liang Wang. Graph contrastive learning with adaptive augmentation. In WWW, 2021

  46. [54]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.