GraphPPD: Posterior Predictive Modelling for Graph-Level Inference

Amine Natik; Liheng Ma; Mark Coates; Soumyasundar Pal; Yingxue Zhang

REVIEW 3 major objections 5 minor 18 references

GraphPPD: Posterior Predictive Modelling for Graph-Level Inference

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A variational module that learns the posterior predictive distribution from labelled context graphs turns any graph encoder into an uncertainty-aware predictor for whole-graph classification and regression.

desk verdict A useful application of attentive neural process ideas to graph-level prediction, but the 'variational PPD' label is unearned and the training protocol has a target/context overlap ambiguity that needs to be resolved before the numbers are fully trustworthy. read the letter →

arxiv 2508.16995 v1 pith:VRYLD3SD submitted 2025-08-23 stat.ML cs.LG

Soumyasundar Pal , Liheng Ma , Amine Natik , Yingxue Zhang , Mark Coates This is my paper · ORCID

classification stat.MLcs.LG

keywords posteriorpredictivedistributiongraph-levelinferenceuncertaintyquantificationgraphneuralnetworksvariationalcross-attentionselectivepredictionamortized

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

The paper sets out to give graph-level prediction tasks—classifying or regressing a property of an entire molecule, protein, or social network—the same kind of uncertainty quantification that node-level methods already enjoy. It proposes a variational framework, GraphPPD, that learns an approximation of the posterior predictive distribution directly from a training set of labelled graphs, in a data-adaptive way. The predictive distribution for a query graph is built by cross-attention over a sampled set of labelled context graphs, so the model can compare the query against similar training examples at inference time. If the framework works as claimed, any existing GNN or graph transformer can be upgraded to produce calibrated predictive distributions without redesigning the encoder, improving accuracy and enabling selective prediction.

What carries the argument

The machinery is the amortized posterior predictive module $q^{\mathrm{emb}}_\phi$, a cross-attention network that maps a target graph embedding and a set of labelled context graph embeddings to the parameters of a predictive distribution. It computes attention coefficients $\alpha_{ij}$ between target query $x_i$ and context key $x_j$, aggregates context values $W_v[x_j \Vert y_j]$ into a representation $r_i$, concatenates this with the target embedding, and passes the result through an MLP to produce either class probabilities or a Gaussian mean and covariance. This object carries the argument because it replaces the intractable integral over model parameters in the definition of the posterior predictive distribution with a single forward pass that is learned end-to-end or on top of a frozen encoder; the Monte Carlo average over context sets in Equation (10) is what turns context sampling into a source of predictive uncertainty.

What would settle it

Take a dataset where test graphs are generated in groups with a shared latent property (for example, molecules from the same synthesis batch with correlated solubility errors), train GraphPPD, and check whether adding the label of one test graph to the context set changes predictive probabilities for the other test graphs in the same group. If the conditional-independence model cannot use that within-group label information to sharpen its predictions, the central claim of a data-adaptive posterior predictive distribution fails on that setting.

Watch

Extended reading notes

Core claim

GraphPPD's central claim is that the posterior predictive distribution for graph-level labels can be learned by amortized variational inference rather than by Bayesian inference over GNN parameters. Conditioning on a context set of labelled graphs, the model factorizes the joint predictive distribution as a product over targets, $q_\lambda(\{y_i\}_{i\in T} \mid \{G_i\}_{i\in T}, D^{\mathrm{graph}}_C) = \prod_{i\in T} q^{\mathrm{emb}}_\phi(y_i \mid x_i, D^{\mathrm{emb}}_C)$, where $x_i$ is the graph encoder's embedding. The factor $q^{\mathrm{emb}}_\phi$ is implemented with cross-attention: each target embedding attends to the context embeddings and labels, producing a context-aware representation that feeds a softmax head for classification or a Gaussian head with learned mean and covariance for regression. At inference, several random context sets are sampled and their predictive distributions are averaged by Monte Carlo. On 7 OGB molecular datasets, 10 TU datasets, and the ZINC-12k regression task, the paper reports that this module improves accuracy, ROC-AUC, or MAE over the underlying encoder (GIN, GINE, GMT, GraphGPS) in most settings, outperforms Monte Carlo dropout, and matches deep ensembles with fewer parameters and less training time.

Load-bearing premise

The load-bearing premise is that, once a batch of labelled context graphs is fixed, the labels of the test graphs are conditionally independent of one another; if test-graph labels carry shared structure beyond what the context set reveals, the learned 'posterior predictive' is misspecified.

Editorial extensions

If this is right

Adding GraphPPD's PPD module to GIN, GINE, GMT, or GraphGPS improves ROC-AUC, accuracy, or MAE on the majority of 18 graph-level benchmarks, with larger relative gains on the OGB molecular datasets (up to 3.93%).
The predictive uncertainty can drive selective prediction: when test graphs with the highest entropy are handed to an oracle reviewer, the collaborative accuracy and ROC-AUC curves of GraphPPD exceed those of the base encoder on datasets such as SIDER, TOXCAST, and TOX21.
The 2-stage variant, which freezes a pretrained encoder and trains only the PPD module on precomputed embeddings, retains most of the improvements while substantially cutting training time, so uncertainty awareness can be added post-hoc to existing GNN pipelines.
GraphPPD is statistically indistinguishable from deep ensembles in Wilcoxon signed-rank tests across the 17 classification datasets, while using fewer trainable parameters and less training time (for GIN on TU datasets, about 39.5% lower training time and 14.5% fewer parameters on average).

Reading between the lines

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

The modular design implies the PPD module can be transplanted to any domain with a strong feature extractor—images, point clouds, or sequences—since the attention module consumes embeddings and labels only; the paper hints at this generality but runs no such experiment.
Relaxing the conditional-independence factorization in Eq. (3) to allow target-to-target attention or a shared latent variable would test whether correlated test labels carry information beyond the context set; this is a natural next step rather than a claim of the paper.
Decomposing the Monte Carlo average in Eq. (10) into context-sampling variance and within-context entropy would give practitioners a direct diagnostic for whether the module's uncertainty reflects genuine data adaptivity or irreducible label noise.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Senior colleague, here's my take. GraphPPD is a useful application paper: it takes the attentive neural process cross-attention idea and puts it on top of a GNN encoder for graph-level classification and regression. The empirical coverage is unusually broad—18 datasets, MC dropout and ensemble comparisons, runtime and parameter counts, plus a selective-prediction experiment. The gains over the base encoders are modest but consistent, and the 2-stage mode (freeze encoder, train PPD module) is a practical selling point for people who already have a pretrained encoder.

The soft spots are about framing and a specification gap, not about the core math. First, the paper calls Eq. (4) a "variational PPD," but there is no variational posterior and no ELBO. What they actually train is a conditional neural process: a cross-attention model that maps a context set to a predictive distribution for target embeddings. That is a legitimate thing, but the variational label is unearned and should be dropped or given a formal link to Eq. (1).

Second, and this is the one that matters: Algorithm 1 samples target and context indices "randomly" from L with no disjointness constraint. Section 4.1 says both are sampled uniformly from the training data. If that is literal, then in training the context can include the target's own graph and label, and the cross-attention in Eq. (6) can copy y_i from W_v[x_i || y_i]. On MUTAG with |T|=128, the expected intersection with a 64-size context is around 43. Even if the model does not always attend to itself, the learning problem is different from test time, where targets are never in the context. The authors might have intended C ~ L\T; if so, they need to say that explicitly. Without a code release or a clarified statement, the reported numbers are not shielded from this failure mode. This is not a fabrication charge; it is an under-specification that a referee has to have resolved before trusting the empirical claims.

Third, the uncertainty evidence is mixed. Table 11 shows ECE and NLL often worse than the plain GNN baseline on the TU datasets. The selective prediction curves in Figure 2 look nice but have no error bars and no competing uncertainty baselines. So the contribution to uncertainty quantification is not convincingly demonstrated.

Overall: this is a paper for people building graph-level predictors who want a lightweight way to add a predictive distribution without paying the cost of ensembles. It deserves peer review, not desk rejection, but it needs revision: fix the variational terminology, specify the disjoint sampling (and ideally release code), and tighten the selective-prediction experiment with error bars and baselines. Set would_accept_peer_review=true.

Referee Report

3 major / 5 minor

Summary. The paper proposes GraphPPD, an amortized module that, given a graph encoder, produces predictive distributions for graph-level classification and regression by cross-attending to a set of labeled context graphs. Training minimizes the negative conditional log-likelihood (Eq. 9) over randomly sampled target and context subsets; inference averages over P sampled context sets (Eq. 10). Experiments compare against the base GNN/transformer encoders, MC dropout, and ensembles on 7 OGB datasets, 10 TU datasets, and ZINC-12k, plus a selective-prediction task. The central claims are that this constitutes a novel variational framework for the posterior predictive distribution and that it yields uncertainty-aware predictions with improved accuracy and selective-prediction performance at lower parameter cost than ensembles.

Significance. If the empirical results are valid, the paper provides a flexible way to attach a context-aware predictive head to existing graph encoders, with a broad evaluation spanning 18 datasets and including parameter/time comparisons against MC dropout and ensembles. The use of a proper-scoring-rule objective (Eq. 9) is a sound starting point for amortized predictive modelling, and the modular design described in Section 4.5 makes the framework easy to extend. However, the contribution as stated is weakened by the unsupported 'variational PPD' label and, more importantly, by an unresolved train-time leakage risk in the sampling protocol of Algorithm 1; the uncertainty-calibration evidence in Appendix A.3 is also mixed and does not uniformly support the uncertainty-awareness claims.

major comments (3)

[Algorithm 1, Section 4.1, Eq. (6)] The training protocol does not require the target set T and context set C to be disjoint: Algorithm 1 line 8 samples both 'randomly' from L, and Section 4.1 states only that both are sampled uniformly from the training data. If a target index i also lies in C, then in the cross-attention computation of Eq. (6) the query x_i can attend to its own associated label y_i via the value vector W_v[x_i || y_i]. The training loss in Eq. (9) can then be minimized by copying labels out of the context rather than by learning a predictive relationship. At inference, targets are test graphs not present in the context set, so the shortcut is unavailable; this asymmetry means the reported gains in Tables 1-5 and Figure 2 could be partly or wholly an artifact of this train-time leakage. The concern is not negligible on small datasets: for example, MUTAG has 188 graphs, and with a target batch of 128 and context size 64 (Table 9), the expected intersection size under uniform sampling without replacement is about 43. The authors must explicitly enforce and state C ∩ T = ∅ (or sample C from L \ T) and rerun the experiments, or provide evidence from the released code that this disjointness is already enforced.
[Section 2, Section 4.1, Section 4.3, Eq. (1) vs Eq. (9)] The paper repeatedly describes the method as a 'variational modelling framework for the posterior predictive distribution', but no variational family, ELBO, KL divergence, or posterior over model parameters is ever defined. Eq. (1) gives the Bayesian posterior predictive as an integral over the parameter posterior p(γ | D_L), while the proposed objective Eq. (9) is an empirical conditional likelihood / proper-scoring-rule objective for the amortized model q_λ(y | G, D_C). No derivation shows that the quantity in Eq. (1) is approximated by Eq. (4). This is a load-bearing mismatch with the abstract and introduction: the contribution can be reframed as an amortized conditional predictive model in the style of neural processes, which is a legitimate contribution, but claiming 'variational' and 'posterior predictive' without a variational derivation overstates the theoretical basis and should be corrected.
[Table 11 (Appendix A.3), Section 6.6] The uncertainty-calibration evidence is mixed and does not consistently support the claim of improved uncertainty quantification. In Table 11, on the ECE criterion the E2E GraphPPD matches or loses to the base encoder in most cases (e.g., GIN ECE win/tie/loss 5/0/5; GMT ECE 3/0/7), and on NLL the results are similar (GIN NLL 5/0/5; GMT NLL 6/1/3). The 2-stage variant has large calibration degradations on several datasets (e.g., D&D ECE 0.175 vs GIN 0.128; PTC NLL 0.644 vs GIN 0.556). The selective-prediction results in Figure 2 compare only against the feature extractor, not against MC dropout or ensembles, and are not accompanied by calibration metrics. The main text's statements about 'effective uncertainty characterization' should be tempered, and the full ECE/NLL/Brier comparisons should be discussed honestly rather than deferred to the appendix.

minor comments (5)

[Algorithm 1] The notation for the test set is confusing: the input line uses L for both the training set D_L and the test set {G_i}_{i ∈ L}, and line 15 says 'For each i ∈ L' where the context suggests the test set. Use distinct symbols, e.g., U for the unlabeled/test set.
[Table 11] The Brier Score rows appear to be identical to the NLL rows for every model and dataset (e.g., GIN Brier 0.416 equals GIN NLL 0.416). This is almost certainly a copy-paste error and must be fixed, since Brier scores for binary/multiclass classification should be computed separately.
[Table 11] The GMT training-time block contains two rows both labeled 'Ours(2-stage)' (roughly 361.3 and 513.4 seconds for different datasets), which is confusing; one of these should presumably be 'Ours(E2E)'.
[Table 9] Several hyperparameter entries are given as slash-separated values, e.g., '1/2', '2/1', '1/4'. It is unclear whether these represent multiple configurations, a range, or different settings for different folds/seeds; please clarify how these values were used.
[Section 4.1, Eq. (3)] The conditional-independence assumption in Eq. (3) is stated without justification. For graph-level tasks with independently drawn test graphs it is plausible, but it should be explicitly identified as a modelling assumption, and the limitation should be discussed for settings where test graphs share latent structure beyond what the context captures.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity of the derivation-chain type; the PPD module is an openly empirical neural-process-style conditional model, trained and tested on held-out benchmarks.

full rationale

The paper does not claim to derive an external quantity from first principles and then feed that same quantity back in. Its central object, q_emb_phi, is explicitly introduced as a parameterized amortized approximation of the posterior predictive density, and equations (3)-(4) are stated modeling choices rather than consequences of equations (1)-(2). The loss in equation (9) is standard negative log-likelihood on targets conditioned on sampled contexts, and the test-time Monte Carlo average in equation (10) is presented as an approximation, not as a theorem forced by a Bayesian posterior. Hyperparameters are tuned on validation subsets and results are reported on held-out test splits, so no fitted parameter is renamed as a prediction. The only self-citations (Zhang et al. 2019, Pal et al. 2020, Sun et al. 2020, Ma et al. 2019) appear in related-work discussion of transductive node/link uncertainty methods and are not load-bearing for the proposed graph-level framework. The possible overlap between target and context sets in Algorithm 1, where both are sampled from the same training set L, is a legitimate empirical leakage concern that could invalidate reported E2E gains if realized, but it is not a circularity of the derivation chain: the claimed reduction would be a training-data artifact, not the paper's equations equating an output with an input. No step in the paper reduces, by construction or by self-citation, to its own inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central machinery is a neural network trained by NLL; the only invented component is the cross-attention PPD module. Free parameters are mostly ML hyperparameters. The main unverified assumptions are the conditional-independence factorization and the implicit claim that the optimized NLL yields calibrated predictive distributions.

free parameters (6)

context size |C| = 32 (ZINC) or 64 (TU/OGB)
Chosen per dataset in hyperparameter search (Table 9); controls the amount of conditioning information and affects performance.
number of attention layers = 1-2 per dataset
Hyperparameter tuned per dataset (Table 9).
number of attention heads = 1-4 per dataset
Hyperparameter tuned per dataset (Table 9).
number of PPD encoder/decoder layers = 1-2 per dataset
Hyperparameter tuned per dataset (Table 9).
number of sampled context sets P at inference = not reported
Equation 10 averages P context sets; P is not specified in the paper, making the inference procedure under-specified.
attention dimension dh = not reported
Dimension of attention keys/queries in Eq. 5 is not given; a free architectural choice.

assumptions (5)

domain assumption Targets are conditionally independent given the context set (Eq. 3)
Standard neural-process assumption; no latent variable or dependence structure is introduced to justify it for graph labels. Used to factor q_λ as a product over targets.
domain assumption The training set D_L is available at inference time and is representative of the test context distribution
Assumed in Section 3 and used in Eq. 10 where context sets are sampled from training data during inference.
domain assumption Minimizing expected cross-entropy in Eq. 9 will produce a calibrated predictive distribution
The paper does not derive any calibration guarantee; the loss is a standard proper scoring rule but calibration requires capacity and coverage assumptions.
domain assumption The chosen GNN/transformer encoders and PPD module have sufficient capacity to represent the true conditional distribution
Invoked implicitly when claiming PPD quality; not verified.
standard math Bayes rule for posterior predictive in Eq. 1-2
Background probability; not the source of the main claim.

invented entities (1)

q_emb PPD module (cross-attention over target/context embeddings)
purpose: Produces an amortized approximation of the posterior predictive distribution for graph-level tasks
The module is a neural network introduced in Section 4.2; no theoretical guarantee that it approximates Eq. 1, and empirical calibration results are mixed (Table 11). The only validation is accuracy/selective-prediction on benchmarks.

how reviews work

0 comments

Cite this review

Pith. "Pith review of GraphPPD: Posterior Predictive Modelling for Graph-Level Inference." pith.science (2026). https://pith.science/paper/VRYLD3SD

@misc{pith2026250816995,
  author       = {Pith},
  title        = {Pith review of: GraphPPD: Posterior Predictive Modelling for Graph-Level Inference},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VRYLD3SD}},
  note         = {Machine review of arXiv:2508.16995}
}

read the original abstract

Accurate modelling and quantification of predictive uncertainty is crucial in deep learning since it allows a model to make safer decisions when the data is ambiguous and facilitates the users' understanding of the model's confidence in its predictions. Along with the tremendously increasing research focus on \emph{graph neural networks} (GNNs) in recent years, there have been numerous techniques which strive to capture the uncertainty in their predictions. However, most of these approaches are specifically designed for node or link-level tasks and cannot be directly applied to graph-level learning problems. In this paper, we propose a novel variational modelling framework for the \emph{posterior predictive distribution}~(PPD) to obtain uncertainty-aware prediction in graph-level learning tasks. Based on a graph-level embedding derived from one of the existing GNNs, our framework can learn the PPD in a data-adaptive fashion. Experimental results on several benchmark datasets exhibit the effectiveness of our approach.

Figures

Figures reproduced from arXiv: 2508.16995 by the authors.

**Figure 1.** Procedure for GraphPPD training and inference through the graph encoder. We use Demb C = {(xj , yj )}j∈C to denote the context representations and their labels. Demb C is subsequently fed to the amortized PPD module along with the target embeddings {xi}i∈T for modelling qλ({yi}i∈T |{Gi}i∈T , Dgraph C ), which is evaluated as: qλ({yi}i∈T |{Gi}i∈T , Dgraph C ) = Y i∈T qλ(yi |Gi , Dgraph C ), (3) = Y i∈T q emb ϕ (yi |x… view at source ↗

**Figure 2.** Selective prediction results on a) PROTEINS, b) REDDIT-M, c) COLLAB, d) SIDER, e) TOXCAST, and f) TOX21 datasets. 6.3. Comparison with Monte Carlo Dropout We compare our approach with Monte Carlo (MC) Dropout (Gal & Ghahramani, 2016), which provides a variational approximation of the posterior of the model parameters. The results in Tables 4 and 5 demonstrate that approximation of the PPD using MC dropout fails t… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references · 11 canonical work pages

[4]

P., Luu, A

Dwivedi, V . P., Luu, A. T., Laurent, T., Bengio, Y ., and Bresson, X. Graph neural networks with learnable structural and positional representations. arXiv preprint arXiv:2110.07875,

arXiv
[5]

J., Ramalho, T., Saxton, D., Shanahan, M., Whye Teh, Y ., Rezende, D

Garnelo, M., Rosenbaum, D., Maddison, C. J., Ramalho, T., Saxton, D., Shanahan, M., Whye Teh, Y ., Rezende, D. J., and Eslami, S. M. A. Conditional Neural Processes. arXiv e-prints, art. arXiv:1807.01613, July 2018a. Garnelo, M., Schwarz, J., Rosenbaum, D., Viola, F., Rezende, D. J., Eslami, S. M. A., and Whye Teh, Y . Neural Processes. arXiv e-prints, ar...

arXiv
[8]

Monti, F., Frasca, F., Eynard, D., Mannion, D., and Bronstein, M. M. Fake news detection on social me- dia using geometric deep learning. arXiv preprint arXiv:1902.06673,

arXiv 1902
[9]

P., Luu, A

Ramp´aˇsek, L., Galkin, M., Dwivedi, V . P., Luu, A. T., Wolf, G., and Beaini, D. Recipe for a general, powerful, scal- able graph transformer. arXiv preprint arXiv:2205.12454,

arXiv
[10]

A framework for recommending accurate and diverse items using Bayesian graph convolutional neural networks

Sun, J., Guo, W., Zhang, D., Zhang, Y ., Regol, F., Hu, Y ., Guo, H., Tang, R., Yuan, H., He, X., and Coates, M. A framework for recommending accurate and diverse items using Bayesian graph convolutional neural networks. In GraphPPD: Posterior Predictive Modelling for Graph-Level Inference Proc. ACM SIGKDD Int. Conf. Knowl. Discov. & Data Mining, pp. 2030...

work page
[11]

The prediction of the quality of results in logic synthesis using transformer and graph neural networks

Yang, C., Xia, Y ., and Chu, Z. The prediction of the quality of results in logic synthesis using transformer and graph neural networks. arXiv preprint arXiv:2207.11437,

work page arXiv
[12]

Summary statistics of the graph datasets (Morris et al., 2020; Hu et al., 2021; Dwivedi et al.,

work page 2020
[13]

TU Datasets D&D PROTEINS MUTAG NCI1 PTC IMDB-B IMDB-M REDDIT-B REDDIT-M COLLAB # Graphs 1178 1113 188 4110 344 1000 1500 2000 5000 5000 Avg. # nodes 284.32 39.06 17.93 29.8 25.5 19.8 13.0 429.6 508.5 74.5 # classes 2 2 2 2 2 2 3 2 5 3 Metric Accuracy OGB Datasets BACE BBBP CLINTOX HIV SIDER TOXCAST TOX21 Regression Dataset ZINC # Graphs 1513 2039 1478 411...

work page 2000

Show all 18 references

[15]

The task is to predict the constrained solubility (logP) of the molecule

ZINC-12K is a molecular dataset with 12K graphs introduced in (Dwivedi et al., 2020), which is the subset of the ZINC-full (Irwin et al., 2012). The task is to predict the constrained solubility (logP) of the molecule. This dataset has a predefined 10K/1K/1K train/validation/t...

2020
[16]

We provide a summary of the hyperparameter configurations in Table

on theRay.Tuneplatform (Liaw et al., 2018). We provide a summary of the hyperparameter configurations in Table

2018
[17]

learnable parameters for OGB datasets

Comparison of ROC-AUC of molecular property prediction, training time, and no. learnable parameters for OGB datasets. Relative increase and no. win/tie/loss is computed w.r.t. the corresponding graph encoder model in each case. Alg. BACE BBBP CLINTOX HIV SIDER TOXCAST TOX21 re...
[18]

Relative increase and no

Comparison of accuracy, training time, and number of learnable parameters for different models for TU datasets. Relative increase and no. win/tie/loss is computed w.r.t. the corresponding graph encoder model in each case. Accuracy D&D PROTEINS MUTAG NCI1 PTC IMDB-B IMDB-M REDD...
[2012]

doi: 10.1021/ci3001277

ISSN 1549-9596. doi: 10.1021/ci3001277. Izmailov, P., Vikram, S., Hoffman, M. D., and Wilson, A. G. What are Bayesian neural network posteriors really like? In Proc. Int. Conf. Machine Learning, Virtual, Jul

doi
[2015]

P., Joshi, C

Dwivedi, V . P., Joshi, C. K., Luu, A. T., Laurent, T., Bengio, Y ., and Bresson, X. Benchmarking graph neural networks. arXiv preprint arXiv:2003.00982,

2003 arXiv
[2017]

E., and Stoica, I

Liaw, R., Liang, E., Nishihara, R., Moritz, P., Gonzalez, J. E., and Stoica, I. Tune: A research platform for dis- tributed model selection and training. arXiv preprint arXiv:1807.05118,

arXiv
[2019]

R., and Li `o, P

Day, B., Cangea, C., Jamasb, A. R., and Li `o, P. Mes- GraphPPD: Posterior Predictive Modelling for Graph-Level Inference sage passing neural processes. ArXiv e-prints: arXiv 2009.13895,

2009 arXiv
[2020]

and Wingate, D

Carr, A. and Wingate, D. Graph neural processes: To- wards Bayesian graph neural networks. ArXiv e-prints, arXiv:1902.10042,

1902 arXiv
[2021]

Many new datasets are included in recent years (Hu et al., 2020)

The open graph benchmark (OGB) is one of the widely used graph benchmarks, initially proposed in (Hu et al., 2021). Many new datasets are included in recent years (Hu et al., 2020). BBBP is related to blood-brain barrier penetration (membrane permeability). TOX21 is from toxic...

2021

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [4]

P., Luu, A

Dwivedi, V . P., Luu, A. T., Laurent, T., Bengio, Y ., and Bresson, X. Graph neural networks with learnable structural and positional representations. arXiv preprint arXiv:2110.07875,

arXiv

[2] [5]

J., Ramalho, T., Saxton, D., Shanahan, M., Whye Teh, Y ., Rezende, D

Garnelo, M., Rosenbaum, D., Maddison, C. J., Ramalho, T., Saxton, D., Shanahan, M., Whye Teh, Y ., Rezende, D. J., and Eslami, S. M. A. Conditional Neural Processes. arXiv e-prints, art. arXiv:1807.01613, July 2018a. Garnelo, M., Schwarz, J., Rosenbaum, D., Viola, F., Rezende, D. J., Eslami, S. M. A., and Whye Teh, Y . Neural Processes. arXiv e-prints, ar...

arXiv

[3] [8]

Monti, F., Frasca, F., Eynard, D., Mannion, D., and Bronstein, M. M. Fake news detection on social me- dia using geometric deep learning. arXiv preprint arXiv:1902.06673,

arXiv 1902

[4] [9]

P., Luu, A

Ramp´aˇsek, L., Galkin, M., Dwivedi, V . P., Luu, A. T., Wolf, G., and Beaini, D. Recipe for a general, powerful, scal- able graph transformer. arXiv preprint arXiv:2205.12454,

arXiv

[5] [10]

A framework for recommending accurate and diverse items using Bayesian graph convolutional neural networks

Sun, J., Guo, W., Zhang, D., Zhang, Y ., Regol, F., Hu, Y ., Guo, H., Tang, R., Yuan, H., He, X., and Coates, M. A framework for recommending accurate and diverse items using Bayesian graph convolutional neural networks. In GraphPPD: Posterior Predictive Modelling for Graph-Level Inference Proc. ACM SIGKDD Int. Conf. Knowl. Discov. & Data Mining, pp. 2030...

work page

[6] [11]

The prediction of the quality of results in logic synthesis using transformer and graph neural networks

Yang, C., Xia, Y ., and Chu, Z. The prediction of the quality of results in logic synthesis using transformer and graph neural networks. arXiv preprint arXiv:2207.11437,

work page arXiv

[7] [12]

Summary statistics of the graph datasets (Morris et al., 2020; Hu et al., 2021; Dwivedi et al.,

work page 2020

[8] [13]

TU Datasets D&D PROTEINS MUTAG NCI1 PTC IMDB-B IMDB-M REDDIT-B REDDIT-M COLLAB # Graphs 1178 1113 188 4110 344 1000 1500 2000 5000 5000 Avg. # nodes 284.32 39.06 17.93 29.8 25.5 19.8 13.0 429.6 508.5 74.5 # classes 2 2 2 2 2 2 3 2 5 3 Metric Accuracy OGB Datasets BACE BBBP CLINTOX HIV SIDER TOXCAST TOX21 Regression Dataset ZINC # Graphs 1513 2039 1478 411...

work page 2000