Pith. sign in

REVIEW 4 major objections 5 minor 105 references

Bridging Domain Adaptation and Graph Neural Networks: A Tensor-Based Framework for Effective Label Propagation

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper claims that label propagation, driven by pseudo-labels, transfers labels across graph domains and that pairing it with a tensor-based GNN encoder outperforms prior domain-adaptive graph classifiers.

desk verdict A sensible and genuinely new integration of pseudo-label-conditioned label propagation with a tensor GNN, but the single-run numbers and uncontrolled protocol leave the SOTA claim unsupported; worth reviewing with major experimental revisions. read the letter →

arxiv 2502.08505 v2 pith:QZPZDTW7 submitted 2025-02-12 cs.LG

Tao Wen , Elynn Chen , Yuzhou Chen , Qi Lei This is my paper · ORCID
classification cs.LG MSC 68T0768R1055N31
keywords graphclassificationunsuperviseddomainadaptationlabelpropagationneuralnetworkspersistenthomologytensorpseudo-labelingtopologicaldataanalysis
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 tries to show that domain adaptation for whole-graph classification can be carried out by label propagation instead of adversarial distribution matching. The proposed LP-TGNN encodes each graph with a tensor-based GNN that combines message-passing layers with persistent-homology topological features, then transfers label information from a labeled source domain to an unlabeled target domain by a consistency regularizer. The regularizer avoids defining a distance between graphs: for each target graph it takes the source graphs whose ground-truth labels agree with the target's pseudo-label as neighbors and pushes the model to agree with them. Across six source-to-target benchmark pairs the method reports 71.8% average accuracy, ahead of 70.5% for the best prior graph-adaptive baselines. A reader should care because it offers a plug-in alternative to adversarial alignment that works with generic GNN encoders.

What carries the argument

The central object is the pseudo-label-conditioned neighbor set of Eq. (6): for a target graph $G^t_j$, the neighbor set $\Pi_j$ consists of source graphs whose ground-truth labels equal the target's pseudo-label $\hat{y}^t_j = \arg\max q^t_j$ from the MLP classifier. This replaces an explicit graph distance, which is hard to define for structured data, with a class-consistency criterion. The consistency regularizer (Eq. 7) is a masked FixMatch-style cross-entropy between the target's prediction and the averaged neighbor-set prediction, applied separately to the convolutional and topological branch outputs, and combined with supervised source loss (Eq. 9). The tensor transformation layer fuses multi-layer GCN outputs and persistence-image tensors while preserving discriminative features via low-rank tensor structure.

What would settle it

Construct a source-target pair or synthetic dataset where, before any adaptation, the MLP's pseudo-labels on target graphs are no better than chance, run LP-TGNN, and check whether the label-propagation term still improves accuracy over the supervised-only model; if it helps, the improvement cannot be attributed to pseudo-label-conditioned neighbor selection as specified in Eq. (6).

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that the bottleneck for domain-adaptive graph classification is not graph distance but representation geometry: once graphs are embedded so that same-class graphs are closer than different-class graphs, label information can propagate across domains through pseudo-labeled neighbor sets. LP-TGNN realizes this by (i) a TGNN encoder whose convolutional branch stacks GCN layers and whose topological branch pools persistence images from multiple filtrations, fused by a tensor transformation layer with low-rank weights, and (ii) a label-propagation objective in which, for each confident target pseudo-label, the source samples of that class are aggregated and used as a consistency target. The paper reports this design outperforms previous adversarial and contrastive graph domain adaptation methods on the standard benchmark transfers and that the same regularizer improves a plain GIN encoder.

Load-bearing premise

The whole mechanism rests on the MLP classifier's pseudo-labels for target graphs being accurate enough that collecting source samples with matching labels actually collects same-class neighbors; if pseudo-labels are noisy, the consistency loss trains the model to agree with its own errors.

Editorial extensions

If this is right

  • If the central claim is right, any GNN encoder, not just TGNN, can be upgraded for unsupervised graph domain adaptation by adding the pseudo-label-conditioned label-propagation regularizer; LP-GIN already matches the previous best average accuracy.
  • Label propagation, being a regularizer, does not force domain-invariant representations, so discriminative target-specific information is retained, unlike adversarial alignment.
  • Both supervised loss and consistency regularization are indispensable: ablations in Table 2 show that removing either term makes results identical across all source domains, i.e., the model stops transferring.
  • The topological branch contributes supplementary signal: removing it hurts performance, but removing the convolutional branch hurts far more, locating representation learning mainly in the GCN branch.
  • Confidence thresholding of pseudo-labels matters: only target samples with max probability above $\tau$ contribute to the consistency loss, so the method trades coverage for pseudo-label quality.

Reading between the lines

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

  • Because the neighbor set is built from pseudo-labels, the method's ceiling is set by the MLP's early accuracy on target graphs; a natural stress test is to measure LP-TGNN's gain as a function of pseudo-label error and to see whether the consistency term degrades performance when pseudo-labels are systematically wrong.
  • The pseudo-label-conditioned neighbor set is close in spirit to a nearest-class-mean or prototype update, so LP-TGNN might be seen as a prototypical consistency method; this connection suggests it could be combined with prototype-based pseudo-label refinement.
  • The paper's synthetic subpopulation-shift study suggests the same regularizer could apply to other structured data with covariate shift, such as point clouds or hypergraphs, whenever a representation with small intra-class distance is available.
  • The paper implicitly claims that adversarial invariance is unnecessary for graph transfer; a direct test would compare LP-TGNN with an adversarial variant on a benchmark where label shift, not covariate shift, dominates, since label propagation assumes shared conditional label distributions.
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 / 5 minor

Summary. The paper proposes LP-TGNN, a framework for unsupervised domain-adaptive graph classification. It combines a tensor-based graph encoder (a GCN branch plus a persistent-homology topological branch fused through a Tensor Transformation Layer) with a label-propagation regularizer. The regularizer, defined in Eq. (6)-(7), builds for each target graph a neighbor set of source graphs whose ground-truth labels match the target graph's pseudo-label, and enforces a FixMatch-style consistency loss. The objective in Eq. (9) sums the supervised source loss and two branch-wise regularization terms. Experiments are reported on six benchmark pairs (Table 1), an ablation on Mutagenicity (Table 2), and a synthetic subpopulation-shift study (Table 3), with pseudo-label quality reported in Table 4.

Significance. If the empirical claims are reliable, LP-TGNN would be a practical and general contribution: the label-propagation formulation avoids adversarial domain alignment, the tensor encoder is a natural fit for combining local and topological graph information, and the reported LP-GIN results suggest the regularizer transfers to other GNN backbones. The authors also release code and include an ablation study and a synthetic experiment. However, the central claim of state-of-the-art accuracy rests on single-run accuracy numbers with no error bars, no significance tests, and no described validation protocol; the average gain over CoCo and LP-GIN is only 1.3 percentage points. The strength of the contribution is therefore currently limited by the experimental evidence rather than by the core idea.

major comments (4)
  1. [§5.1, Table 1] The central empirical claim that LP-TGNN 'outperforms baselines by a notable margin' is not supported by the reported protocol. Section 5.1 states that the learning rate is chosen from {0.01, 0.05, 0.1} and that the threshold tau is fixed at 0.8, but it never describes a validation split or the rule used to select these hyperparameters. If the per-task learning rate was selected using target test accuracy, the comparison is leakage-inflated. In addition, all numbers in Table 1 appear to be single-run accuracies: there are no error bars, no repeated-seed statistics, and no significance tests. Since LP-TGNN's average advantage over CoCo and LP-GIN is only 1.3 points, the reported gain is within the range that tuning noise or a favorable seed could produce. The authors should provide multi-seed results with standard deviations, a clear model-selection rule, and a statistical comparison (e.g., paired tests across the six tasks).
  2. [Table 2, §5.3] The ablation table contains degenerate patterns that the paper acknowledges but does not explain. For example, in the LP-TGNN/LP row the accuracy for every source-to-M0 task is exactly 44.8, every source-to-M2 task is 62.7, and every source-to-M3 task is 46.9; the LP-TGNN/Sup row shows the same target-only repetition for M0, M2, and M3 targets. The text states that 'for either method, the results on the same target domain are identical for any source domain' and interprets this as loss of transferability, but such exact repetition across different source domains is not what one would expect from a model that still trains on the source data. This suggests either a bug in the ablation implementation, a degenerate solution (e.g., predicting the target class prior), or an evaluation artifact. Because the full method's advantage over the ablations is used to justify both the supervised loss and the label-propagation regularizer, the authors must diagnose these identical values and rerun the ablations before the component analysis can be accepted.
  3. [§4.5, Eq. (6)-(7)] The neighbor set Pi_j is defined as the source samples in the current mini-batch whose ground-truth label equals the target pseudo-label. When a mini-batch contains no source sample of that class, Pi_j is empty and the averaging operation Z^s_j = AVG({Z^s_i}_{i in Pi_j}) is undefined. The paper does not describe how empty neighbor sets are handled, even though with random mini-batches and six-class Mutagenicity tasks this will occur. This is a concrete algorithmic gap that affects training stability and reproducibility. The authors should specify the fallback behavior (e.g., skip the term, use a smaller batch, or maintain a memory bank) and report whether results change.
  4. [§4.5, Table 4] The label-propagation regularizer conditions on the MLP's own pseudo-labels, creating a self-training loop that can amplify errors. Table 4 reports pseudo-label accuracy and classification accuracy but does not analyze how pseudo-label errors propagate into the regularization objective or the final classifier. The correlation claim is descriptive and does not establish that the pseudo-labeling step is robust. The authors should provide an analysis (e.g., oracle-label upper bound, or accuracy as a function of confidence threshold tau) to show that the method is not primarily benefiting from a favorable error pattern.
minor comments (5)
  1. [§4.3, Eq. (4)] The normalization in the definition of A-hat appears to be a typo: it reads \tilde D^{-1/2} \tilde A \tilde D^{1/2}, which is not the standard symmetric normalization; it should likely be \tilde D^{-1/2} \tilde A \tilde D^{-1/2}. Please correct.
  2. [§5.4, Table 3] The synthetic study is described in only three sentences. The construction of the subpopulation shift (e.g., how the 1:2 and 2:1 ratios were enforced, which class was shifted, and how many runs were averaged) should be specified for reproducibility, and the table would benefit from repeated-seed statistics.
  3. [Figure 1] The figure caption contains typographical errors: 'Non-Eyzyme' should be 'Non-Enzyme' and 'Enzyme versus Non-Eyzyme' appears in the main text as well.
  4. [References] Reference [92] duplicates reference [28]; one of the two entries should be removed or replaced with a distinct citation.
  5. [§5.3, Table 2] The caption for Table 2 is missing an abbreviation explanation: the meaning of 'Topo', 'Conv', 'Sup', and 'LP' in the row labels should be stated in the caption or the surrounding text for readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical SOTA claim is evaluated on held-out target labels and is not forced by construction.

full rationale

The paper's central claim is empirical (Table 1 accuracy on held-out target labels), not a derivation, and no predicted quantity is defined in terms of its own fitted inputs. The label propagation regularizer (Eqs. 6-9) does use the model's own pseudo-labels to define source neighbor sets, which is a self-training-style feedback loop, but the reported classification accuracy is measured against ground-truth target labels hidden during training, so the result is not equal to the pseudo-labels or to any fitted parameter by construction. The TGNN encoder is taken from the authors' prior work [28], but that is an independently published architecture and is used as a component rather than as a proof step; the paper's experiments benchmark the combined method on external datasets. The theoretical support for label propagation is cited from [26], which includes a co-author, but that is a published, parameter-free theoretical result with stated assumptions that do not include the present method, and it serves as motivation rather than as the source of the reported numbers. Concerns about the lack of a validation split, uncontrolled learning-rate selection, and absent repeated-seed variance are experimental-protocol or correctness risks, not circularity, because they do not make any claimed result identical to its input by definition. Overall, the derivation chain is self-contained with respect to the empirical claims, and any self-citations are not load-bearing in a circular sense.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim rests on inherited assumptions (covariate shift, transfer of label propagation to graphs, embedding separability, low-rank tensor structure) rather than new theoretical justifications. No new entities are postulated.

free parameters (4)
  • confidence threshold tau = 0.8
    Target samples are included in the label propagation regularizer only when max softmax probability exceeds 0.8; this value is set by hand in Section 5.1 and is load-bearing for the pseudo-label filter.
  • learning rate = selected from {0.01, 0.05, 0.1}
    The implementation section lists this set but does not report the chosen value or the validation procedure, so the final model may be selected using target accuracy.
  • persistence image resolution = 50 x 50
    PI resolution for the topological branch is fixed at 50x50 without ablation or sensitivity analysis.
  • number of filtration functions = 4
    Four centrality filtrations (degree, betweenness, eigenvector, closeness) are chosen by hand for the topological branch.
assumptions (4)
  • domain assumption Source and target domains satisfy covariate shift: PDs(X) != PDt(X) and PDs(y|X) = PDt(y|X)
    Stated in Section 3 and used to justify the domain adaptation objective.
  • domain assumption Label propagation theory from [26] transfers to graph spaces
    The paper applies the consistency regularization derived for vector data to graph representations without a theoretical extension; validation is only empirical.
  • domain assumption TGNN embeddings keep intra-class distances below inter-class distances, making pseudo-label propagation reliable
    Figure 1 and Section 1 rely on this geometric separability for the neighbor-set definition; only t-SNE visualizations support it.
  • domain assumption The tensor low-rank inductive bias is suitable for the concatenated GCN and persistence-image tensors
    The TTL layer is taken from [28]; the paper does not analyze whether the low-rank assumption holds for these graph-derived tensors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Domain Adaptation and Graph Neural Networks: A Tensor-Based Framework for Effective Label Propagation." pith.science (2026). https://pith.science/paper/QZPZDTW7

@misc{pith2026250208505,
  author       = {Pith},
  title        = {Pith review of: Bridging Domain Adaptation and Graph Neural Networks: A Tensor-Based Framework for Effective Label Propagation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QZPZDTW7}},
  note         = {Machine review of arXiv:2502.08505}
}
read the original abstract

Graph Neural Networks (GNNs) have recently become the predominant tools for studying graph data. Despite state-of-the-art performance on graph classification tasks, GNNs are overwhelmingly trained in a single domain under supervision, thus necessitating a prohibitively high demand for labels and resulting in poorly transferable representations. To address this challenge, we propose the Label-Propagation Tensor Graph Neural Network (LP-TGNN) framework to bridge the gap between graph data and traditional domain adaptation methods. It extracts graph topological information holistically with a tensor architecture and then reduces domain discrepancy through label propagation. It is readily compatible with general GNNs and domain adaptation techniques with minimal adjustment through pseudo-labeling. Experiments on various real-world benchmarks show that our LP-TGNN outperforms baselines by a notable margin. We also validate and analyze each component of the proposed framework in the ablation study.

Figures

Figures reproduced from arXiv: 2502.08505 by the authors.

Figure 1
Figure 1. A visualization of domain adaptation on graph classification. The red line is the deci￾sion boundary of an Enzyme versus Non-Eyzyme classifier trained on SE and SNE. The labels can transfer from the source domain to the target do￾main via methods such as label propagation [26]. For effective label propagation between domains, the intra-class distance is supposed to be smaller than the inter-class distance [27]. For … view at source ↗
Figure 2
Figure 2. An illustration of representations produced by the TTG-NN and the state-of-the-art GIN [ [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The architecture of the proposed LP-TGNN. The graph mini-batches from both domains are [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

105 extracted references · 64 canonical work pages

  1. [1]

    Tudataset: A collection of benchmark datasets for learning with graphs

    ChristopherMorris,NilsM.Kriege,FrankaBause,KristianKersting,PetraMutzel,andMarion Neumann. Tudataset: A collection of benchmark datasets for learning with graphs. In ICML 2020 Workshop on Graph Representation Learning and Beyond (GRL+ 2020), 2020. URL www.graphlearning.io

    2020

  2. [2]

    How powerful are graph neural networks? InProceedings of International Conference on Learning Representations, 2018

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks? InProceedings of International Conference on Learning Representations, 2018

    2018

  3. [3]

    Hierarchical graph representation learning with differentiable pooling

    Zhitao Ying, Jiaxuan You, Christopher Morris, Xiang Ren, Will Hamilton, and Jure Leskovec. Hierarchical graph representation learning with differentiable pooling. InAdvances in Neural Information Processing Systems, volume 31, 2018

    2018

  4. [4]

    kGCN: a graph-based deep learning framework for chemical structures.Journal of Cheminformatics, 12(1), 5 2020

    Ryosuke Kojima, Shoichi Ishida, Masateru Ohta, Hiroaki Iwata, Teruki Honma, and Yasushi Okuno. kGCN: a graph-based deep learning framework for chemical structures.Journal of Cheminformatics, 12(1), 5 2020. doi: 10.1186/s13321-020-00435-6. URLhttps://jcheminf. biomedcentral.com/articles/10.1186/s13321-020-00435-6

    work page doi:10.1186/s13321-020-00435-6 2020

  5. [5]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolu- tional networks. InInternational Conference on Learning Representations, 2017. URL https: //openreview.net/forum?id=SJU4ayYgl

    2017

  6. [6]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. Advances in neural information processing systems, 30, 2017

    2017

  7. [7]

    Improving graph neural networks with learn- able propagation operators

    Moshe Eliasof, Lars Ruthotto, and Eran Treister. Improving graph neural networks with learn- able propagation operators. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors,Proceedings of the 40th International Conference on Machine Learning, volume 202 ofProceedings of Machine Learning Resear...

  8. [8]

    Agent-based graph neural networks

    Karolis Martinkus, Pál András Papp, Benedikt Schesch, and Roger Wattenhofer. Agent-based graph neural networks. InThe Eleventh International Conference on Learning Representations,

Show all 105 references
  1. [9]

    Schoenholz, Patrick F

    Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. Neural message passing for quantum chemistry. In Doina Precup and Yee Whye Teh, editors, Proceedings of the 34th International Conference on Machine Learning, volume 70 of Proceedings of ...

    2017

  2. [10]

    Graphattentionnetworks

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and YoshuaBengio. Graphattentionnetworks. ProceedingsoftheInternationalConferenceonLearning Representations, 2018. 10

    2018

  3. [11]

    Asgn: An active semi-supervised graph neural network for molec- ular property prediction

    Zhongkai Hao, Chengqiang Lu, Zhenya Huang, Hao Wang, Zheyuan Hu, Qi Liu, Enhong Chen, and Cheekong Lee. Asgn: An active semi-supervised graph neural network for molec- ular property prediction. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discover...

    2020

  4. [12]

    Randomized schur complement views for graph contrastive learning

    Vignesh Kothapalli. Randomized schur complement views for graph contrastive learning. In International Conference on Machine Learning, 2023

    2023

  5. [13]

    SEGA: Structural entropy guided anchor view for graph contrastive learning

    Junran Wu, Xueyuan Chen, Bowen Shi, Shangzhe Li, and Ke Xu. SEGA: Structural entropy guided anchor view for graph contrastive learning. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, editors,Proceedings of the 40th In...

  6. [14]

    Transferable attention for domain adaptation.Proceedings of the AAAI Conference on Artificial Intelligence, 33(01): 5345–5352, Jul

    Ximei Wang, Liang Li, Weirui Ye, Mingsheng Long, and Jianmin Wang. Transferable attention for domain adaptation.Proceedings of the AAAI Conference on Artificial Intelligence, 33(01): 5345–5352, Jul. 2019. doi: 10.1609/aaai.v33i01.33015345. URLhttps://ojs.aaai.org/index. php/AA...

    2019 doi

  7. [15]

    Domainadaptation in remote sensing image classification: A survey.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15:9842–9859, 2022

    JiangtaoPeng, YiHuang, WeiweiSun, NaChen, YujieNing, andQianDu. Domainadaptation in remote sensing image classification: A survey.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15:9842–9859, 2022. doi: 10.1109/JSTARS.2022.3220875

    2022

  8. [16]

    Bridging domains with approximately shared features

    Ziliang Samuel Zhong, Xiang Pan, and Qi Lei. Bridging domains with approximately shared features. arXiv preprint arXiv:2403.06424, 2024

    2024 arXiv

  9. [17]

    Near-optimal linear regression under distribution shift

    Qi Lei, Wei Hu, and Jason Lee. Near-optimal linear regression under distribution shift. In International Conference on Machine Learning, pages 6164–6174. PMLR, 2021

    2021

  10. [18]

    Controllable prompt tuning for balancing group distributional robustness.arXiv preprint arXiv:2403.02695, 2024

    Hoang Phan, Andrew Gordon Wilson, and Qi Lei. Controllable prompt tuning for balancing group distributional robustness.arXiv preprint arXiv:2403.02695, 2024

    2024 arXiv

  11. [19]

    Unsupervised domain adaptation for semantic image segmentation: a comprehensive survey, 2021

    Gabriela Csurka, Riccardo Volpi, and Boris Chidlovskii. Unsupervised domain adaptation for semantic image segmentation: a comprehensive survey, 2021

    2021

  12. [20]

    Graph convolutional policy network for goal-directed molecular graph generation.Advances in neural information processing systems, 31, 2018

    Jiaxuan You, Bowen Liu, Zhitao Ying, Vijay Pande, and Jure Leskovec. Graph convolutional policy network for goal-directed molecular graph generation.Advances in neural information processing systems, 31, 2018

    2018

  13. [21]

    Skipgnn: predicting molecular interactions with skip-graph networks.Scientific reports, 10(1):1–16, 2020

    Kexin Huang, Cao Xiao, Lucas M Glass, Marinka Zitnik, and Jimeng Sun. Skipgnn: predicting molecular interactions with skip-graph networks.Scientific reports, 10(1):1–16, 2020

    2020

  14. [22]

    Does gnn pretraining help molecular representa- tion? Advances in Neural Information Processing Systems, 35:12096–12109, 2022

    Ruoxi Sun, Hanjun Dai, and Adams Wei Yu. Does gnn pretraining help molecular representa- tion? Advances in Neural Information Processing Systems, 35:12096–12109, 2022

    2022

  15. [23]

    Graph domain adaptation via theory-grounded spectral regularization

    Yuning You, Tianlong Chen, Zhangyang Wang, and Yang Shen. Graph domain adaptation via theory-grounded spectral regularization. InThe Eleventh International Conference on Learning Representations, 2023. URLhttps://openreview.net/forum?id=OysfLgrk8mk

    2023

  16. [24]

    Domain-adversarial training of neural net- works.JournalofMachineLearningResearch ,17(59):1–35,2016

    Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, Mario March, and Victor Lempitsky. Domain-adversarial training of neural net- works.JournalofMachineLearningResearch ,17(59):1–35,2016. URL http://jmlr.org/papers/ v17/15-239.html

    2016

  17. [25]

    Tuan Nguyen, Toan Tran, Yarin Gal, Philip Torr, and Atilim Gunes Baydin

    A. Tuan Nguyen, Toan Tran, Yarin Gal, Philip Torr, and Atilim Gunes Baydin. KL guided domain adaptation. InInternational Conference on Learning Representations, 2022. URLhttps: //openreview.net/forum?id=0JzqUlIVVDd. 11

    2022

  18. [26]

    A theory of label propagation for subpopulation shift

    Tianle Cai, Ruiqi Gao, Jason Lee, and Qi Lei. A theory of label propagation for subpopulation shift. In Marina Meila and Tong Zhang, editors,Proceedings of the 38th International Conference on Machine Learning, volume 139 ofProceedings of Machine Learning Research, pages 1170–...

    2021

  19. [27]

    HaoChen, Colin Wei, Ananya Kumar, and Tengyu Ma

    Jeff Z. HaoChen, Colin Wei, Ananya Kumar, and Tengyu Ma. Beyond separability: An- alyzing the linear transferability of contrastive representations to related subpopulations. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors,Ad- vances in Neural Inf...

    2022

  20. [28]

    Tensor-view topological graph neural network

    Tao Wen, Elynn Chen, and Yuzhou Chen. Tensor-view topological graph neural network. In SanjoyDasgupta,StephanMandt,andYingzhenLi,editors, ProceedingsofThe27thInternational Conference on Artificial Intelligence and Statistics, volume 238 ofProceedings of Machine Learning Resear...

    2024

  21. [29]

    Structure- activity relationship of mutagenic aromatic and heteroaromatic nitro compounds

    A K Debnath, R L Lopez de Compadre, G Debnath, A J Shusterman, and C Hansch. Structure- activity relationship of mutagenic aromatic and heteroaromatic nitro compounds. correlation with molecular orbital energies and hydrophobicity.J. Med. Chem., 34(2):786–797, February 1991

    1991

  22. [30]

    Visualizing data using t-sne

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of Machine Learning Research, 9(86):2579–2605, 2008. URL http://jmlr.org/papers/v9/ vandermaaten08a.html

    2008

  23. [31]

    Conditionaladversarial domain adaptation

    MingshengLong,ZhangjieCao,JianminWang,andMichaelI.Jordan. Conditionaladversarial domain adaptation. InProceedings of the 32nd International Conference on Neural Information Processing Systems, NIPS’18, page 1647–1657, Red Hook, NY, USA, 2018. Curran Associates Inc

    2018

  24. [32]

    Graph neural networks: A review of methods and applications

    Jie Zhou, Ganqu Cui, Shengding Hu, Zhengyan Zhang, Cheng Yang, Zhiyuan Liu, Lifeng Wang, Changcheng Li, and Maosong Sun. Graph neural networks: A review of methods and applications. AI Open, 1:57–81, 2020

    2020

  25. [33]

    Weisfeiler-lehman graph kernels.Journal of Machine Learning Research, 12(9), 2011

    Nino Shervashidze, Pascal Schweitzer, Erik Jan Van Leeuwen, Kurt Mehlhorn, and Karsten M Borgwardt. Weisfeiler-lehman graph kernels.Journal of Machine Learning Research, 12(9), 2011

    2011

  26. [34]

    Graph learning: A survey.IEEE Trans AI, 2(2):109–127, 2021

    Feng Xia, Ke Sun, Shuo Yu, Abdul Aziz, Liangtian Wan, Shirui Pan, and Huan Liu. Graph learning: A survey.IEEE Trans AI, 2(2):109–127, 2021

    2021

  27. [35]

    Semi-supervised classification with graph convolutional networks

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. Proceedings of the International Conference on Learning Representations, 2017

    2017

  28. [36]

    Graph attention networks.Proceedings of International Conference on Learning Representations, 2018

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. Graph attention networks.Proceedings of International Conference on Learning Representations, 2018

    2018

  29. [37]

    Towards sparse hierarchical graph classifiers.Workshop on Relational Representation Learning, NeurIPS, 2018

    Cătălina Cangea, Petar Veličković, Nikola Jovanović, Thomas Kipf, and Pietro Liò. Towards sparse hierarchical graph classifiers.Workshop on Relational Representation Learning, NeurIPS, 2018

    2018

  30. [38]

    Graph u-nets

    Hongyang Gao and Shuiwang Ji. Graph u-nets. InProceedings of the International Conference on Machine Learning, pages 2083–2092, 2019

    2019

  31. [39]

    M. Horn, E. De Brouwer, M. Moor, Y. Moreau, B. Rieck, and K. Borgwardt. Topological graph neural networks. InProceedings of International Conference on Learning Representations, 2021. 12

    2021

  32. [40]

    Carrière, F

    M. Carrière, F. Chazal, Y. Ike, T. Lacombe, M. Royer, and Y. Umeda. Perslay: A neural network layer for persistence diagrams and new graph topological signatures. InProceedings of the International Conference on Artificial Intelligence and Statistics, pages 2786–2796, 2020

    2020

  33. [41]

    Going beyond persistent homology using persistenthomology

    Johanna Immonen, Amauri Souza, and Vikas Garg. Going beyond persistent homology using persistenthomology. In37thConferenceonNeuralInformationProcessingSystems(NeurIPS) ,2023. URL https://nips.cc/. Conference on Neural Information Processing Systems, NeurIPS ; Conference date: ...

    2023

  34. [42]

    Adapting visual category models to new domains

    Kate Saenko, Brian Kulis, Mario Fritz, and Trevor Darrell. Adapting visual category models to new domains. In Kostas Daniilidis, Petros Maragos, and Nikos Paragios, editors,Computer Vision–ECCV2010 ,pages213–226,Berlin,Heidelberg,2010.SpringerBerlinHeidelberg. ISBN 978-3-642-15561-1

    2010

  35. [43]

    Unsupervised domain adaptation by backpropagation

    Yaroslav Ganin and Victor Lempitsky. Unsupervised domain adaptation by backpropagation. In Francis Bach and David Blei, editors,Proceedings of the 32nd International Conference on MachineLearning, volume37of ProceedingsofMachineLearningResearch , pages1180–1189, Lille, France,...

    2015

  36. [44]

    Exploring the landscape of distributional robustness for question answering models, 2022

    Anas Awadalla, Mitchell Wortsman, Gabriel Ilharco, Sewon Min, Ian Magnusson, Hannaneh Hajishirzi, and Ludwig Schmidt. Exploring the landscape of distributional robustness for question answering models, 2022

    2022

  37. [45]

    The effect of natural distribution shift on question answering models

    John Miller, Karl Krauth, Benjamin Recht, and Ludwig Schmidt. The effect of natural distribution shift on question answering models. In Hal Daumé III and Aarti Singh, ed- itors, Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of ...

    2020

  38. [46]

    Multi-source unsu- pervised domain adaptation via pseudo target domain.IEEE Transactions on Image Pro- cessing, 31:2122–2135, 2022

    Chuan-Xian Ren, Yong-Hui Liu, Xi-Wen Zhang, and Ke-Kun Huang. Multi-source unsu- pervised domain adaptation via pseudo target domain.IEEE Transactions on Image Pro- cessing, 31:2122–2135, 2022. ISSN 1941-0042. doi: 10.1109/tip.2022.3152052. URL http: //dx.doi.org/10.1109/TIP.2...

    2022

  39. [47]

    Curriculum graphco-teachingformulti-targetdomainadaptation.In ProceedingsoftheIEEE/CVFConference on Computer Vision and Pattern Recognition, 2021

    Subhankar Roy, Evgeny Krivosheev, Zhun Zhong, Nicu Sebe, and Elisa Ricci. Curriculum graphco-teachingformulti-targetdomainadaptation.In ProceedingsoftheIEEE/CVFConference on Computer Vision and Pattern Recognition, 2021

    2021

  40. [48]

    Deep domain confusion: Maximizing for domain invariance, 2014

    Eric Tzeng, Judy Hoffman, Ning Zhang, Kate Saenko, and Trevor Darrell. Deep domain confusion: Maximizing for domain invariance, 2014

    2014

  41. [49]

    Learningtransferablefeatures with deep adaptation networks

    MingshengLong, YueCao, JianminWang, andMichaelJordan. Learningtransferablefeatures with deep adaptation networks. In Francis Bach and David Blei, editors,Proceedings of the 32nd International Conference on Machine Learning, volume 37 ofProceedings of Machine Learning Research,...

    2015

  42. [50]

    Domain-adversarial neural networks, 2015

    Hana Ajakan, Pascal Germain, Hugo Larochelle, François Laviolette, and Mario Marchand. Domain-adversarial neural networks, 2015

    2015

  43. [51]

    Pseudo-label: Thesimpleandefficientsemi-supervisedlearningmethodfor deep neural networks.ICML 2013 Workshop : Challenges in Representation Learning (WREPL), 07 2013

    Dong-HyunLee. Pseudo-label: Thesimpleandefficientsemi-supervisedlearningmethodfor deep neural networks.ICML 2013 Workshop : Challenges in Representation Learning (WREPL), 07 2013

    2013

  44. [52]

    Structural re-weighting improves graph domain adaptation

    Shikun Liu, Tianchun Li, Yongbin Feng, Nhan Tran, Han Zhao, Qiang Qiu, and Pan Li. Structural re-weighting improves graph domain adaptation. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jonathan Scarlett, ed- itors, Proceedings of the...

    2023

  45. [53]

    Dane: Domain adaptive networkembedding

    Yizhou Zhang, Guojie Song, Lun Du, Shuwen Yang, and Yilun Jin. Dane: Domain adaptive networkembedding. InProceedingsoftheTwenty-EighthInternationalJointConferenceonArtificial Intelligence,IJCAI-19,pages4362–4368.InternationalJointConferencesonArtificialIntelligence Organizatio...

    2019 doi

  46. [54]

    Unsuperviseddomain adaptive graph convolutional networks

    ManWu, ShiruiPan, ChuanZhou, XiaojunChang, andXingquanZhu. Unsuperviseddomain adaptive graph convolutional networks. InProceedings of The Web Conference 2020, WWW ’20, page 1457–1467, New York, NY, USA, 2020. Association for Computing Machinery. ISBN 9781450370233. doi: 10.114...

    2020

  47. [55]

    Shift-robust gnns: Over- coming the limitations of localized graph training data

    Qi Zhu, Natalia Ponomareva, Jiawei Han, and Bryan Perozzi. Shift-robust gnns: Over- coming the limitations of localized graph training data. In M. Ranzato, A. Beygelz- imer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan, editors, Advances in Neu- ral Information Processing Sy...

    2021

  48. [56]

    Deal: An unsupervised domain adaptive framework for graph-level classification

    Nan Yin, Li Shen, Baopu Li, Mengzhu Wang, Xiao Luo, Chong Chen, Zhigang Luo, and Xian- Sheng Hua. Deal: An unsupervised domain adaptive framework for graph-level classification. In Proceedings of the 30th ACM International Conference on Multimedia, MM ’22, page 3470–3479, New ...

    2022

  49. [57]

    CoCo: A coupled contrastive framework for unsupervised domain adaptive graph classification

    Nan Yin, Li Shen, Mengzhu Wang, Long Lan, Zeyu Ma, Chong Chen, Xian-Sheng Hua, and Xiao Luo. CoCo: A coupled contrastive framework for unsupervised domain adaptive graph classification. In Andreas Krause, Emma Brunskill, Kyunghyun Cho, Barbara Engelhardt, Sivan Sabato, and Jon...

    2023

  50. [58]

    Domain adaptive graph classifi- cation, 2023

    Siyang Luo, Ziyi Jiang, Zhenghan Chen, and Xiaoxuan Liang. Domain adaptive graph classifi- cation, 2023. URLhttps://arxiv.org/abs/2312.13536

    2023 arXiv

  51. [59]

    Improving predictive inference under covariate shift by weighting the log-likelihood function.Journal of Statistical Planning and Inference, 90(2):227–244, 2000

    Hidetoshi Shimodaira. Improving predictive inference under covariate shift by weighting the log-likelihood function.Journal of Statistical Planning and Inference, 90(2):227–244, 2000. ISSN 0378-3758. doi: https://doi.org/10.1016/S0378-3758(00)00115-4. URLhttps://www. sciencedi...

    doi

  52. [60]

    Fixmatch: Simplifying semi- supervised learning with consistency and confidence

    Kihyuk Sohn, David Berthelot, Nicholas Carlini, Zizhao Zhang, Han Zhang, Colin A Raf- fel, Ekin Dogus Cubuk, Alexey Kurakin, and Chun-Liang Li. Fixmatch: Simplifying semi- supervised learning with consistency and confidence. In H. Larochelle, M. Ranzato, R. Had- sell, M.F. Bal...

    2020

  53. [61]

    Borgwardt

    Nino Shervashidze, Pascal Schweitzer, Erik Jan van Leeuwen, Kurt Mehlhorn, and Karsten M. Borgwardt. Weisfeiler-lehman graph kernels.Journal of Machine Learning Research, 12(77): 2539–2561, 2011. URLhttp://jmlr.org/papers/v12/shervashidze11a.html

    2011

  54. [62]

    Bronstein

    Cristian Bodnar, Fabrizio Frasca, Nina Otter, Yu Guang Wang, Pietro Liò, Guido Montufar, and Michael M. Bronstein. Weisfeiler and lehman go cellular: CW networks. In A. Beygelz- imer, Y. Dauphin, P. Liang, and J. Wortman Vaughan, editors,Advances in Neural Information Processi...

    2021

  55. [63]

    Accurate learning of graph representations with graph multiset pooling

    Jinheon Baek, Minki Kang, and Sung Ju Hwang. Accurate learning of graph representations with graph multiset pooling. InInternational Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=JHcqXGaqiGn

    2021

  56. [64]

    Toalign: Task- oriented alignment for unsupervised domain adaptation

    Guoqiang Wei, Cuiling Lan, Wenjun Zeng, Zhizheng Zhang, and Zhibo Chen. Toalign: Task- oriented alignment for unsupervised domain adaptation. In A. Beygelzimer, Y. Dauphin, P. Liang, and J. Wortman Vaughan, editors,Advances in Neural Information Processing Systems,

  57. [65]

    Metaalign: Coordinating do- main alignment and classification for unsupervised domain adaptation

    Guoqiang Wei, Cuiling Lan, Wenjun Zeng, and Zhibo Chen. Metaalign: Coordinating do- main alignment and classification for unsupervised domain adaptation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16643–16653, June 2021

    2021

  58. [66]

    Thenormmustgoon: Dynamic unsupervised domain adaptation by normalization

    M.JehanzebMirza, JakubMicorek, HorstPossegger, andHorstBischof. Thenormmustgoon: Dynamic unsupervised domain adaptation by normalization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14765–14775, June 2022

    2022

  59. [67]

    Statistical inference for high-dimensional matrix-variate factor models

    Elynn Chen and Jianqing Fan. Statistical inference for high-dimensional matrix-variate factor models. Journal of the American Statistical Association, 118(542):1038–1055, 2023

    2023

  60. [68]

    On projection robust optimal transport: Sample complexity and model misspecification

    Tianyi Lin, Zeyu Zheng, Elynn Chen, Marco Cuturi, and Michael I Jordan. On projection robust optimal transport: Sample complexity and model misspecification. InAISTATS, 2021, pages 262–270. PMLR, 2021

    2021

  61. [69]

    Identification and estimation of threshold matrix-variate factor models

    Xialu Liu and Elynn Chen. Identification and estimation of threshold matrix-variate factor models. Scandinavian Journal of Statistics, 2022

    2022

  62. [70]

    Elynn Chen, Dong Xia, Chencheng Cai, and Jianqing Fan. Semi-parametric tensor factor analysis by iteratively projected singular value decomposition.Journal of the Royal Statistical Society Series B: Statistical Methodology, page qkae001, 2024

    2024

  63. [71]

    Time-varyingmatrixfactormodels, 2024

    BinChen,ElynnChen,StevensonBolivar,andRongChen. Time-varyingmatrixfactormodels, 2024

    2024

  64. [72]

    Computation of optimal mev in decentralized exchanges

    Mengqian Zhang, Yuhao Li, Xinyuan Sun, Elynn Chen, and Xi Chen. Computation of optimal mev in decentralized exchanges. Technical report, Working paper-https://mengqian-zhang. github. io/papers/batch. pdf, Tech. Rep, 2024

    2024

  65. [73]

    Mev makes ev- eryonehappyundergreedysequencingrule

    Yuhao Li, Mengqian Zhang, Jichen Li, Elynn Chen, Xi Chen, and Xiaotie Deng. Mev makes ev- eryonehappyundergreedysequencingrule. In Proceedingsofthe2023WorkshoponDecentralized Finance and Security, pages 9–15, 2023

    2023

  66. [74]

    Data-driven knowledge transfer in batchQ∗ learning

    Elynn Chen, Xi Chen, and Wenbo Jing. Data-driven knowledge transfer in batchQ∗ learning. arXiv preprint arXiv:2404.15209, 2024, 2024

    2024

  67. [75]

    Dynamic contextual pricing with doubly non- parametric random utility models.arXiv preprint arXiv:2405.06866, 2024

    Elynn Chen, Xi Chen, Lan Gao, and Jiayu Li. Dynamic contextual pricing with doubly non- parametric random utility models.arXiv preprint arXiv:2405.06866, 2024

    2024 arXiv

  68. [76]

    Community network auto-regression for high- dimensional time series.Journal of Econometrics, 235(2):1239–1256, 2023

    Elynn Chen, Jianqing Fan, and Xuening Zhu. Community network auto-regression for high- dimensional time series.Journal of Econometrics, 235(2):1239–1256, 2023

    2023

  69. [77]

    Constrained factor models for high-dimensional matrix-variate time series.Journal of the American Statistical Association, 2019

    Elynn Chen, Ruey S Tsay, and Rong Chen. Constrained factor models for high-dimensional matrix-variate time series.Journal of the American Statistical Association, 2019

    2019

  70. [78]

    Modeling multivariate spatial-temporal data with latent low-dimensional dynamics.arXiv preprint arXiv:2002.01305, 2020

    Elynn Chen, Xin Yun, Rong Chen, and Qiwei Yao. Modeling multivariate spatial-temporal data with latent low-dimensional dynamics.arXiv preprint arXiv:2002.01305, 2020

    2002 arXiv

  71. [79]

    Modeling dynamic transport network with matrix factor models: with an application to international trade flow.Journal of Data Science, 2022

    Elynn Chen and Rong Chen. Modeling dynamic transport network with matrix factor models: with an application to international trade flow.Journal of Data Science, 2022. 15

    2022

  72. [80]

    Transfer q-learning.arXiv preprint arXiv:2202.04709, 2022

    Elynn Chen, Sai Li, and Michael I Jordan. Transfer q-learning.arXiv preprint arXiv:2202.04709, 2022

    2022

  73. [81]

    Reinforcement learning in latent heterogeneous environments

    Elynn Chen, Rui Song, and Michael I Jordan. Reinforcement learning in latent heterogeneous environments. Journal of the American Statistical Association, (just-accepted):1–32, 2024

    2024

  74. [82]

    Distributedtensorprincipalcomponent analysis with data heterogneity.arXiv preprint arXiv:2405.11681, 2024

    ElynnChen,XiChen,WenboJing,andYichenZhang. Distributedtensorprincipalcomponent analysis with data heterogneity.arXiv preprint arXiv:2405.11681, 2024

    2024 arXiv

  75. [83]

    Advancing information integration throughempiricallikelihood: Selectivereviewsandanewidea

    Chixiang Chen, Jia Liang, Elynn Chen, and Ming Wang. Advancing information integration throughempiricallikelihood: Selectivereviewsandanewidea. arXivpreprintarXiv:2407.00561 , 2024

    2024 arXiv

  76. [84]

    High-dimensional tensor classification with cp low-rank discriminant structure.arXiv preprint arXiv:2409.14397, 2024

    Elynn Chen, Yuefeng Han, and Jiayu Li. High-dimensional tensor classification with cp low-rank discriminant structure.arXiv preprint arXiv:2409.14397, 2024

    2024 arXiv

  77. [85]

    High-dimensional tensor discriminant analysis with incomplete tensors.arXiv preprint arXiv:2410.14783, 2024

    Elynn Chen, Yuefeng Han, and Jiayu Li. High-dimensional tensor discriminant analysis with incomplete tensors.arXiv preprint arXiv:2410.14783, 2024

    2024 arXiv

  78. [86]

    Stochastic linear bandits with latent hetero- geneity

    Elynn Chen, Xi Chen, Wenbo Jing, and Xiao Liu. Stochastic linear bandits with latent hetero- geneity. arXiv preprint arXiv:2502.00423, 2025

    2025

  79. [87]

    Statistical inference for low-rank tensor models.arXiv preprint arXiv:2501.16223, 2025

    Ke Xu, Elynn Chen, and Yuefeng Han. Statistical inference for low-rank tensor models.arXiv preprint arXiv:2501.16223, 2025

    2025 arXiv

  80. [88]

    On the expressive power of deep learning: A tensor analysis

    Nadav Cohen, Or Sharir, and Amnon Shashua. On the expressive power of deep learning: A tensor analysis. InConference on Learning Theory, pages 698–728, 2016

    2016

  81. [89]

    Tensor contraction layers for parsimonious deep nets.CVPR, pages 1940–1946, 2017

    JeanKossaifi,AranKhanna,ZacharyC.Lipton,TommasoFurlanello,andAnimaAnandkumar. Tensor contraction layers for parsimonious deep nets.CVPR, pages 1940–1946, 2017

    1940

  82. [90]

    Lipton, Arinbjorn Kolbeinsson, Aran Khanna, Tommaso Furlanello, and Anima Anandkumar

    Jean Kossaifi, Zachary C. Lipton, Arinbjorn Kolbeinsson, Aran Khanna, Tommaso Furlanello, and Anima Anandkumar. Tensor regression networks.JMLR, 21:1–21, 2020

    2020

  83. [91]

    Graph tensor networks: An intuitive framework for designing large-scale neural learning systems on multiple domains.arXiv preprint arXiv:2303.13565, 2023

    Yao Lei Xu, Kriton Konstantinidis, and Danilo P Mandic. Graph tensor networks: An intuitive framework for designing large-scale neural learning systems on multiple domains.arXiv preprint arXiv:2303.13565, 2023

    2023 arXiv

  84. [92]

    Tensor-view topological graph neural network

    Tao Wen, Elynn Chen, and Yuzhou Chen. Tensor-view topological graph neural network. In AISTATS, 2024, València SPAIN, 2024

    2024

  85. [93]

    Tensor-fusedmulti-viewgraphcontrastive learning

    YujiaWu,JunyiMo,ElynnChen,andYuzhouChen. Tensor-fusedmulti-viewgraphcontrastive learning. arXiv preprint arXiv:2410.15247, 2024

    2024 arXiv

  86. [94]

    Conditional prediction roc bands for graph classification.AISTATS, 2025, Mai Khao, Thailand, 2024

    Yujia Wu, Bo Yang, Elynn Chen, Yuzhou Chen, and Zheshi Zheng. Conditional prediction roc bands for graph classification.AISTATS, 2025, Mai Khao, Thailand, 2024

    2025

  87. [95]

    Condi- tional uncertainty quantification for tensorized topological neural networks.arXiv preprint arXiv:2410.15241, 2024

    Yujia Wu, Bo Yang, Yang Zhao, Elynn Chen, Yuzhou Chen, and Zheshi Zheng. Condi- tional uncertainty quantification for tensorized topological neural networks.arXiv preprint arXiv:2410.15241, 2024

    2024 arXiv

  88. [96]

    Teaformers: Tensor-augmented transformers for multi-dimensional time series forecasting.arXiv preprint arXiv:2410.20439, 2024

    LinghangKong,ElynnChen,YuzhouChen,andYuefengHan. Teaformers: Tensor-augmented transformers for multi-dimensional time series forecasting.arXiv preprint arXiv:2410.20439, 2024

    2024 arXiv

  89. [97]

    Persistence images: A stable vector representation of persistent homology.Journal of Machine Learning Research, 18, 2017

    HenryAdams,TeganEmerson,MichaelKirby,RachelNeville,ChrisPeterson,PatrickShipman, Sofya Chepushtanova, Eric Hanson, Francis Motta, and Lori Ziegelmeier. Persistence images: A stable vector representation of persistent homology.Journal of Machine Learning Research, 18, 2017. 16

    2017

  90. [98]

    Derivation and validation of toxicophores for mutagenicity prediction.J

    Jeroen Kazius, Ross McGuire, and Roberta Bursi. Derivation and validation of toxicophores for mutagenicity prediction.J. Med. Chem., 48(1):312–320, January 2005

    2005

  91. [99]

    Dobson and Andrew J

    Paul D. Dobson and Andrew J. Doig. Distinguishing enzyme structures from non-enzymes without alignments.Journal of Molecular Biology, 330(4):771–783, 2003. ISSN 0022-2836. doi: https://doi.org/10.1016/S0022-2836(03)00628-4. URL https://www.sciencedirect.com/ science/article/pi...

    2003 doi

  92. [100]

    Spline-Fitting with a genetic algorithm: A method for developing classification {Structure–Activity} relationships.J

    Jeffrey J Sutherland, Lee A O’Brien, and Donald F Weaver. Spline-Fitting with a genetic algorithm: A method for developing classification {Structure–Activity} relationships.J. Chem. Inf. Comput. Sci., 43(6):1906–1915, November 2003. A. Appendix A.1. Tensor-Augmented Neural Net...

    1906

  93. [103]

    • GCN [5] follows the message-passing framework to update node representations itera- tively with neighboring nodes

    Graph learning methods: • WLSubtree[ 61]presentsafamilyofefficientgraphkernelsusingtheWeisfeiler-Lehman test to measure the similarity of graphs. • GCN [5] follows the message-passing framework to update node representations itera- tively with neighboring nodes. • GIN[2]isasta...

  94. [104]

    • ToAlign[64]decomposessourcedomainfeaturesintotask-relatedfeaturesforalignment and task-irrelevant features to be avoided, based on classification meta-knowledge

    Domain alignment methods: • CDAN [31] proposes an adversarial learning framework and conditions on the discrim- inative information from classifier predictions. • ToAlign[64]decomposessourcedomainfeaturesintotask-relatedfeaturesforalignment and task-irrelevant features to be a...

  95. [105]

    • CoCo [57] consists of coupled branches for graph representation learning and con- trastive learning between branches and domains

    Domain adaptive graph classification methods: • DEAL[56]utilizesadversariallearningandadaptiveperturbationfordomainalignment and distillation for pseudo-labeling. • CoCo [57] consists of coupled branches for graph representation learning and con- trastive learning between bran...

  96. [2021]

    URL https://openreview.net/forum?id=XP9SZpjZkq

  97. [2023]

    URL https://openreview.net/forum?id=8WTAh0tj2jC

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.