REVIEW 2 major objections 5 minor 45 references
Instance-Prototype Affinity Learning for Non-Exemplar Continual Graph Learning
T0 review · 2 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read IPAL claims that prototype contrastive learning drifts less than prototype replay, and builds a non-exemplar continual graph learner on that finding.
desk verdict Solid empirical paper with a flawed theory appendix: IPAL works on the benchmarks, but Theorem 1 should be a conjecture, not a theorem. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The carrying object is the PCL objective, a prototypical contrastive loss that treats each node embedding as a query, its own class prototype as the positive, and all other class prototypes as negatives. IPAL makes three modifications to it: TIGP replaces mean prototypes with PageRank-weighted Gaussian distributions, biasing class geometry toward influential nodes; IPAD distills the inner products between current-task instances and old-class prototypes, preserving relational structure rather than rigid feature distances; and DBP selects the top-entropy instances near class boundaries and adds them as extra negative pairs. Together these pieces carry the argument that old-class geometry can be protected without raw exemplars.
What would settle it
Compute the mean feature of each old class before and after training on a new task under PCL and under standard prototype replay; if PCL's mean shift is not consistently smaller across task transitions, the theorem on feature drift is not supported by the data.
Extended reading notes
Core claim
The paper's central discovery is that replacing the cross-entropy objective of prototype replay with a contrastive objective aligns current-task instances with their class prototypes and actively repels instances from other classes, so old feature distributions move less when the encoder is updated. Theorem 1 formalizes this: under Gaussian feature distributions, the KL divergence between previous-task feature distributions before and after an update is smaller for PCL than for Prototype Replay. IPAL operationalizes the result with Topology-Integrated Gaussian Prototypes (TIGP), which weight class means and covariances by PageRank; Instance-Prototype Affinity Distillation (IPAD), which regularizes instance-prototype relationship patterns instead of exact features; and Decision Boundary Perception (DBP), which adds high-entropy boundary instances as negative pairs. In the reported experiments, IPAL has the highest average accuracy on all four benchmarks, such as 83.07 percent on CS-CL versus 81.19 percent for the strongest rehearsal-based comparison.
Load-bearing premise
The load-bearing premise is that the contrastive objective stabilizes old-class feature means, an effect the paper asserts in its proof rather than derives from the loss.
Editorial extensions
If this is right
- If PCL drifts less than Prototype Replay, non-exemplar continual graph learners can reduce their reliance on post-hoc drift-compensation machinery.
- IPAL's benchmark results suggest that relational distillation plus contrastive prototypes can substitute for feature distillation, which the paper argues over-constrains plasticity.
- Topology-weighted prototypes imply that node importance affects prototype quality, so uniform averaging of node features is suboptimal for graph continual learning.
- Boundary-aware negative pairs imply that explicitly mining hard instances improves inter-class separation during incremental graph learning.
- A better plasticity-stability trade-off means graph classifiers can be trained on sequentially arriving classes while keeping no raw data from past tasks.
- This is an editorial inference: if the drift-reduction claim holds, the same contrastive-prototype mechanism may transfer to other exemplar-free continual learning settings, such as images or text, where stored prototypes are also affected by feature drift.
- This is an editorial inference: a quantitative drift metric, measuring the mean shift of old-class features before and after each task, would make the paper's central premise directly testable; the paper currently supports it with a qualitative visualization.
- This is an editorial inference: the PageRank weighting suggests a testable family of variants that replace PageRank with other centrality measures and compare plasticity, or that recompute prototype weights as the graph evolves.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes IPAL, a non-exemplar continual graph learning method built on prototype contrastive learning (PCL). The authors argue empirically and via a stated theorem (Theorem 1, Appendix A) that PCL induces less feature drift than conventional prototype replay (PR), and they introduce three components: Topology-Integrated Gaussian Prototypes (TIGP), which PageRank-weight node contributions when computing class prototypes; Instance-Prototype Affinity Distillation (IPAD), which regularizes instance–prototype affinity via Mixup-based synthetic features; and Decision Boundary Perception (DBP), which adds hard-example repulsion to the PCL objective. The method is evaluated on CS-CL, CoraFull-CL, Arxiv-CL, and Reddit-CL against regularization, rehearsal, and non-exemplar baselines, reporting improved average performance in Table 1, with ablations and parameter sensitivity studies in Tables 2 and Figure 6.
Significance. If the claims hold, IPAL is a competitive empirical contribution to non-exemplar continual graph learning, a setting where memory and privacy constraints rule out raw-example replay. The paper's strengths are the standard CGLB-style benchmark setup, the inclusion of several strong non-exemplar baselines (POLO, EFC), the clear ablation isolating TIGP, IPAD, and DBP, and the consistent average-performance gains in Table 1. However, the central scientific claim that PCL provably incurs less feature drift than PR is not supported by the proof in Appendix A; the key step is asserted rather than derived. Because the main text explicitly presents Theorem 1 as a rigorous proof that motivates the entire design, this gap is load-bearing and must be repaired, or the theoretical claim must be removed and reframed as an empirical observation.
major comments (2)
- [Appendix A, Theorem 1 and Eq. (21)] Theorem 1 is not established by the proof given. The pivotal step is the sentence after Eq. (21): "Eq. 2 regularizes the negative sample gradients in Eq. 17 by minimizing F_θt(x_t)⊤·µ_m, encouraging ∆θ to drive F_θt(x_m) closer to F_θt−1(x_m)." Eq. (17) is a gradient with respect to the current-task feature f_θt = F_θt(x_t), whereas Eq. (21) is an expectation over previous-task samples (x_m, y_m) ∈ T_{0:t−1}, which are not available when training on T_t. No argument shows that optimizing the PCL objective on current-task samples contracts the previous-task prototype means or bounds ||µ′_m − µ_m||². The subsequent covariance claim that PR "allows Σ_m to expand uncontrollably" is qualitative and is not derived from the PR objective in Eq. (1). Consequently, the inequality D_KL(P^{t−1}∥P^t)_{PCL} < D_KL(P^{t−1}∥P^t)_{PR} does not follow from the equations presented. Because Section 1 presents this theorem as the rigorous justification for the PCL-based design, this gap needs to be repaired or the theorem must be removed and the motivation reframed as an empirical observation.
- [Section 5 / Figure 1] The empirical premise that "PCL exhibits less pronounced drift than conventional PR" is currently supported only by the qualitative visualization in Figure 1. No quantitative drift metric is reported for any dataset or task, such as the average prototype mean shift or an estimated KL divergence between the old and new feature distributions. Since Theorem 1 is the only formal support for this premise and that proof is incomplete, the paper should either report such a metric (e.g., on CS-CL and CoraFull-CL) or explicitly downgrade the claim to an observed trend that motivates the method rather than a demonstrated property of PCL.
minor comments (5)
- [Algorithm 1, lines 9–10] As typeset, the drift compensation and the TIGP generation/allocation appear inside the "for epoch" loop, so they would be executed every epoch and the memory buffer would be rewritten multiple times; move them outside the epoch loop to match the post-task descriptions in Sections 4.1 and 4.4.
- [Eq. (7)] The notation "idx = arg max ... == y_m" is unclear: y_m is not defined for the prototype index m, and the comparison with the arg max result should be spelled out (e.g., by defining the prototype label and how the boolean mask is applied).
- [Section 5.1, Implementation Details] "PyTorch 3.10 framework" should read "PyTorch with Python 3.10" (or equivalent); PyTorch does not have a version 3.10.
- [Section 4.4, Eq. (11)] The text says β is "assigned an exceedingly small value" in light of Figure 1, but the experiments use β = 0.1; please clarify whether 0.1 is considered small in this context or whether the reported results use a different value.
- [Figure 3] The caption describes top/middle/bottom rows, but the figure contains multiple dataset columns; please annotate each subplot with its method and dataset, or arrange the panels by method, to remove ambiguity about which heatmap corresponds to which configuration.
Circularity Check
Theorem 1's proof assumes the conclusion: PCL is asserted to 'essentially regularize' the exact mean-shift quantity the theorem must bound, so the theoretical motivation is circular, while the empirical benchmarks remain independent.
-
other
[Appendix A, proof of Theorem 1 (text after Eq. 21 and after Eq. 23)]
"Eq. 2 regularizes the negative sample gradients in Eq. 17 by minimizing Fθt(xt)⊤·µm, encouraging ∆θ to drive Fθt(xm) closer to Fθt−1(xm). Thus, LPCL essentially regularizes ∥µ′m−µm∥2 2, effectively alleviating feature shift. ... According to our preceding analysis, PCL encourages alignment between Fθt(xm) and Fθt−1(xm), whereas PR, lacking such a constraint, allows Σm to expand uncontrollably."
Theorem 1 aims to prove DKL(Pt−1∥Pt)PCL < DKL(Pt−1∥Pt)PR, which the proof reduces to showing PCL yields smaller ∥µ′m−µm∥2 and tr(Σ′−1m Σm−I). The mean-shift step asserts, without derivation, that minimizing the negative term Fθt(xt)⊤·µm for current-task samples drives previous-task features Fθt(xm) toward Fθt−1(xm), then declares LPCL 'essentially regularizes' the exact norm ∥µ′m−µm∥22. Equation 21 is an expectation over previous-task samples (xm,ym), while Eq. 17's gradient is evaluated on current-task samples (xt,yt); no transfer or bound connects them. The covariance step explicitly invokes 'our preceding analysis', i.e., the same unproved alignment claim. The theorem's conclusion is therefore assumed rather than derived from the PCL objective.
full rationale
The benchmark comparison and ablations are self-contained: Table 1 reports standard train/test accuracy on held-out nodes, γ is tuned by grid search, and ablation variants are trained and evaluated under the same protocol, so no fitted parameter is renamed as a prediction. The circularity is confined to the theoretical motivation. The proof of Theorem 1 (Appendix A) needs to show PCL produces smaller mean shift and covariance expansion than PR; instead, after writing Eq. 21 it asserts that PCL 'essentially regularizes ∥µ′m−µm∥22', which is exactly the target inequality, and the covariance argument simply repeats 'our preceding analysis'. This is a circular step because the property to be proven is inserted as an interpretation of Eq. 2 without a mathematical derivation. The Figure 1 support is a visualization rather than a quantitative drift metric, which compounds the gap but is not itself circular. This flaw does not falsify the empirical SOTA claim, which stands on independent evaluation.
Assumptions & free parameters
free parameters (5)
- gamma =
0.6 to 0.7 on small datasets, higher on larger (grid search over 0.1-1.0)
- beta =
0.1
- K =
10
- S_t subset size =
100
- Mixup lambda distribution =
Beta(9,21), lambda in [0,0.4]
assumptions (4)
- domain assumption Feature distributions for each class and the drifted distributions are Gaussian (N(µ,Σ))
- ad hoc to paper PCL's negative-sample gradient keeps prior prototype means close to their old values
- domain assumption PageRank value reflects node importance for prototype computation
- domain assumption Task graph can be decomposed into disjoint subgraphs by class labels
Cite this review
Pith. "Pith review of Instance-Prototype Affinity Learning for Non-Exemplar Continual Graph Learning." pith.science (2026). https://pith.science/paper/AG2JAH3O
@misc{pith2026250510040,
author = {Pith},
title = {Pith review of: Instance-Prototype Affinity Learning for Non-Exemplar Continual Graph Learning},
year = {2026},
howpublished = {\url{https://pith.science/paper/AG2JAH3O}},
note = {Machine review of arXiv:2505.10040}
}
read the original abstract
Graph Neural Networks (GNN) endure catastrophic forgetting, undermining their capacity to preserve previously acquired knowledge amid the assimilation of novel information. Rehearsal-based techniques revisit historical examples, adopted as a principal strategy to alleviate this phenomenon. However, memory explosion and privacy infringements impose significant constraints on their utility. Non-Exemplar methods circumvent the prior issues through Prototype Replay (PR), yet feature drift presents new challenges. In this paper, our empirical findings reveal that Prototype Contrastive Learning (PCL) exhibits less pronounced drift than conventional PR. Drawing upon PCL, we propose Instance-Prototype Affinity Learning (IPAL), a novel paradigm for Non-Exemplar Continual Graph Learning (NECGL). Exploiting graph structural information, we formulate Topology-Integrated Gaussian Prototypes (TIGP), guiding feature distributions towards high-impact nodes to augment the model's capacity for assimilating new knowledge. Instance-Prototype Affinity Distillation (IPAD) safeguards task memory by regularizing discontinuities in class relationships. Moreover, we embed a Decision Boundary Perception (DBP) mechanism within PCL, fostering greater inter-class discriminability. Evaluations on four node classification benchmark datasets demonstrate that our method outperforms existing state-of-the-art methods, achieving a better trade-off between plasticity and stability.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
Memory aware synapses: Learning what (not) to forget
Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuyte- laars. Memory aware synapses: Learning what (not) to forget. InProceedings of the European conference on computer vision (ECCV), pages 139–154, 2018
2018
-
[2]
Elahe Arani, Fahad Sarfraz, and Bahram Zonooz. Learning fast, learning slow: A general contin- ual learning method based on complementary learning system.arXiv preprint arXiv:2201.12604, 2022
arXiv 2022
-
[3]
Mm-gnn: Mix-moment graph neural network towards modeling neighborhood feature distribution
Wendong Bi, Lun Du, Qiang Fu, Yanlin Wang, Shi Han, and Dongmei Zhang. Mm-gnn: Mix-moment graph neural network towards modeling neighborhood feature distribution. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining, pages 132–140, 2023
work page 2023
-
[4]
De Cheng, Yuxin Zhao, Nannan Wang, Guozhang Li, Dingwen Zhang, and Xinbo Gao. Efficient statistical sampling adaptation for exemplar-free class incremental learning.IEEE Transactions on Circuits and Systems for Video Technology, 2024
work page 2024
-
[5]
Protognn: Prototype-assisted message passing framework for non-homophilous graphs
Yanfei Dong, Mohammed Haroon Dupty, Lambert Deng, Yong Liang Goh, and Wee Sun Lee. Protognn: Prototype-assisted message passing framework for non-homophilous graphs. 2022
work page 2022
-
[6]
Exemplar-free continual representation learning via learnable drift compensation
Alex Gomez-Villa, Dipam Goswami, Kai Wang, Andrew D Bagdanov, Bartlomiej Twardowski, and Joost van de Weijer. Exemplar-free continual representation learning via learnable drift compensation. InEuropean Conference on Computer Vision, pages 473–490. Springer, 2024
work page 2024
-
[7]
Inductive representation learning on large graphs.Advances in neural information processing systems, 30, 2017
Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs.Advances in neural information processing systems, 30, 2017
2017
-
[8]
Contrastive multi-view representation learning on graphs
Kaveh Hassani and Amir Hosein Khasahmadi. Contrastive multi-view representation learning on graphs. InInternational conference on machine learning, pages 4116–4126. PMLR, 2020
work page 2020
Show all 45 references
-
[9]
Distilling the knowledge in a neural network
Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015
2015 arXiv
-
[10]
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. Advances in neural information processing systems, 33:22118–22133, 2020
2020
-
[11]
Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016
Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016
2016 arXiv
-
[12]
Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13):3521–3526, 2017
James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of scienc...
2017
-
[13]
Fcs: Feature calibration and separation for non- exemplar class incremental learning
Qiwei Li, Yuxin Peng, and Jiahuan Zhou. Fcs: Feature calibration and separation for non- exemplar class incremental learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 28495–28504, 2024
2024
-
[14]
Inductive graph few-shot class incremental learning
Yayong Li, Peyman Moghadam, Can Peng, Nan Ye, and Piotr Koniusz. Inductive graph few-shot class incremental learning. InProceedings of the Eighteenth ACM International Conference on Web Search and Data Mining, pages 466–474, 2025. 10
2025
-
[15]
Learning without forgetting.IEEE transactions on pattern analysis and machine intelligence, 40(12):2935–2947, 2017
Zhizhong Li and Derek Hoiem. Learning without forgetting.IEEE transactions on pattern analysis and machine intelligence, 40(12):2935–2947, 2017
2017
-
[16]
Overcoming catastrophic forgetting in graph neural networks
Huihui Liu, Yiding Yang, and Xinchao Wang. Overcoming catastrophic forgetting in graph neural networks. InProceedings of the AAAI conference on artificial intelligence, volume 35, pages 8653–8661, 2021
2021
-
[17]
Cat: Balanced continual graph learning with graph condensation
Yilun Liu, Ruihong Qiu, and Zi Huang. Cat: Balanced continual graph learning with graph condensation. In2023 IEEE International Conference on Data Mining (ICDM), pages 1157–
-
[18]
Gradient episodic memory for continual learning
David Lopez-Paz and Marc’Aurelio Ranzato. Gradient episodic memory for continual learning. Advances in neural information processing systems, 30, 2017
2017
-
[19]
Elastic feature consolidation for cold start exemplar-free incremental learning.arXiv preprint arXiv:2402.03917, 2024
Simone Magistri, Tomaso Trinci, Albin Soutif-Cormerais, Joost van de Weijer, and Andrew D Bagdanov. Elastic feature consolidation for cold start exemplar-free incremental learning.arXiv preprint arXiv:2402.03917, 2024
2024 arXiv
-
[20]
Automating the construction of internet portals with machine learning.Information Retrieval, 3:127–163, 2000
Andrew Kachites McCallum, Kamal Nigam, Jason Rennie, and Kristie Seymore. Automating the construction of internet portals with machine learning.Information Retrieval, 3:127–163, 2000
2000
-
[21]
Distributed repre- sentations of words and phrases and their compositionality.Advances in neural information processing systems, 26, 2013
Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. Distributed repre- sentations of words and phrases and their compositionality.Advances in neural information processing systems, 26, 2013
2013
-
[22]
Replay-and-forget-free graph class-incremental learning: A task profiling and prompting approach.arXiv preprint arXiv:2410.10341, 2024
Chaoxi Niu, Guansong Pang, Ling Chen, and Bing Liu. Replay-and-forget-free graph class-incremental learning: A task profiling and prompting approach.arXiv preprint arXiv:2410.10341, 2024
2024 arXiv
-
[23]
The pagerank citation ranking: Bringing order to the web
Lawrence Page, Sergey Brin, Rajeev Motwani, and Terry Winograd. The pagerank citation ranking: Bringing order to the web. Technical report, Stanford infolab, 1999
1999
-
[24]
Fetril: Feature translation for exemplar-free class-incremental learning
Grégoire Petit, Adrian Popescu, Hugo Schindler, David Picard, and Bertrand Delezoide. Fetril: Feature translation for exemplar-free class-incremental learning. InProceedings of the IEEE/CVF winter conference on applications of computer vision, pages 3911–3920, 2023
2023
-
[25]
Incremental graph classification by class prototype construction and augmentation
Yixin Ren, Li Ke, Dong Li, Hui Xue, Zhao Li, and Shuigeng Zhou. Incremental graph classification by class prototype construction and augmentation. InProceedings of the 32nd ACM International Conference on Information and Knowledge Management, pages 2136–2145, 2023
2023
-
[26]
Graph neural networks: Architectures, stability, and transferability.Proceedings of the IEEE, 109(5):660–682, 2021
Luana Ruiz, Fernando Gama, and Alejandro Ribeiro. Graph neural networks: Architectures, stability, and transferability.Proceedings of the IEEE, 109(5):660–682, 2021
2021
-
[27]
The lattice theory of information.Transactions of the IRE professional Group on Information Theory, 1(1):105–107, 1953
Claude Shannon. The lattice theory of information.Transactions of the IRE professional Group on Information Theory, 1(1):105–107, 1953
1953
-
[28]
A mathematical theory of communication.The Bell system technical journal, 27(3):379–423, 1948
Claude E Shannon. A mathematical theory of communication.The Bell system technical journal, 27(3):379–423, 1948
1948
-
[29]
Pitfalls of graph neural network evaluation.arXiv preprint arXiv:1811.05868, 2018
Oleksandr Shchur, Maximilian Mumme, Aleksandar Bojchevski, and Stephan Günnemann. Pitfalls of graph neural network evaluation.arXiv preprint arXiv:1811.05868, 2018
2018 arXiv
-
[30]
Prototypical networks for few-shot learning
Jake Snell, Kevin Swersky, and Richard Zemel. Prototypical networks for few-shot learning. Advances in neural information processing systems, 30, 2017
2017
-
[31]
Graph attention networks.arXiv preprint arXiv:1710.10903, 2017
Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks.arXiv preprint arXiv:1710.10903, 2017
2017 arXiv
-
[32]
Non-exemplar class- incremental learning via adaptive old class reconstruction
Shaokun Wang, Weiwei Shi, Yuhang He, Yifan Yu, and Yihong Gong. Non-exemplar class- incremental learning via adaptive old class reconstruction. InProceedings of the 31st ACM International Conference on Multimedia, pages 4524–4534, 2023. 11
2023
-
[33]
Mixup for node and graph classification
Yiwei Wang, Wei Wang, Yuxuan Liang, Yujun Cai, and Bryan Hooi. Mixup for node and graph classification. InProceedings of the Web Conference 2021, pages 3663–3674, 2021
2021
-
[34]
Simplifying graph convolutional networks
Felix Wu, Amauri Souza, Tianyi Zhang, Christopher Fifty, Tao Yu, and Kilian Weinberger. Simplifying graph convolutional networks. InInternational conference on machine learning, pages 6861–6871. Pmlr, 2019
2019
-
[35]
Infogcl: Information-aware graph contrastive learning.Advances in Neural Information Processing Systems, 34:30414–30425, 2021
Dongkuan Xu, Wei Cheng, Dongsheng Luo, Haifeng Chen, and Xiang Zhang. Infogcl: Information-aware graph contrastive learning.Advances in Neural Information Processing Systems, 34:30414–30425, 2021
2021
-
[36]
How powerful are graph neural networks?arXiv preprint arXiv:1810.00826, 2018
Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks?arXiv preprint arXiv:1810.00826, 2018
2018 arXiv
-
[37]
Semantic drift compensation for class-incremental learning
Lu Yu, Bartlomiej Twardowski, Xialei Liu, Luis Herranz, Kai Wang, Yongmei Cheng, Shangling Jui, and Joost van de Weijer. Semantic drift compensation for class-incremental learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6982...
2020
-
[38]
Continual learning through synaptic intelligence
Friedemann Zenke, Ben Poole, and Surya Ganguli. Continual learning through synaptic intelligence. InInternational conference on machine learning, pages 3987–3995. PMLR, 2017
2017
-
[39]
Fine-grained knowledge selection and restoration for non-exemplar class incremental learning
Jiang-Tian Zhai, Xialei Liu, Lu Yu, and Ming-Ming Cheng. Fine-grained knowledge selection and restoration for non-exemplar class incremental learning. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 6971–6978, 2024
2024
-
[40]
mixup: Beyond empirical risk minimization.arXiv preprint arXiv:1710.09412, 2017
Hongyi Zhang, Moustapha Cisse, Yann N Dauphin, and David Lopez-Paz. mixup: Beyond empirical risk minimization.arXiv preprint arXiv:1710.09412, 2017
2017 arXiv
-
[41]
Continual learning on dynamic graphs via parameter isolation
Peiyan Zhang, Yuchen Yan, Chaozhuo Li, Senzhang Wang, Xing Xie, Guojie Song, and Sunghun Kim. Continual learning on dynamic graphs via parameter isolation. InProceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages ...
2023
-
[42]
Cglb: Benchmark tasks for continual graph learning.Advances in Neural Information Processing Systems, 35:13006–13021, 2022
Xikun Zhang, Dongjin Song, and Dacheng Tao. Cglb: Benchmark tasks for continual graph learning.Advances in Neural Information Processing Systems, 35:13006–13021, 2022
2022
-
[43]
Sparsified subgraph memory for continual graph representation learning
Xikun Zhang, Dongjin Song, and Dacheng Tao. Sparsified subgraph memory for continual graph representation learning. In2022 IEEE International Conference on Data Mining (ICDM), pages 1335–1340. IEEE, 2022
2022
-
[44]
Overcoming catastrophic forgetting in graph neural networks with experience replay
Fan Zhou and Chengtai Cao. Overcoming catastrophic forgetting in graph neural networks with experience replay. InProceedings of the AAAI conference on artificial intelligence, volume 35, pages 4714–4722, 2021
2021
-
[45]
Prototype augmentation and self-supervision for incremental learning
Fei Zhu, Xu-Yao Zhang, Chuang Wang, Fei Yin, and Cheng-Lin Liu. Prototype augmentation and self-supervision for incremental learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5871–5880, 2021. A Theoretical Analysis on Feature D...
2021
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.