Prototype Compensation Method for Small-Scale Incremental Learning Based on Attention-Guided Sample Augmentation and Causal Constraints

Jiamin Zhi; Kexin Zhang; Meiqi Zhao

doi:10.54097/wjejtx49

Authors

Jiamin Zhi SIAS UNIVERSITY FACULTY OF ENGINEERING Zhengzhou, Henan 451100, China
Kexin Zhang SIAS UNIVERSITY FACULTY OF ENGINEERING Zhengzhou, Henan 451100, China
Meiqi Zhao SIAS UNIVERSITY FACULTY OF ENGINEERING Zhengzhou, Henan 451100, China

DOI:

https://doi.org/10.54097/wjejtx49

Keywords:

Small-start non-replay incremental learning; catastrophic forgetting; attention-guided sample augmentation; causal constraints; prototype drift compensation; stability-plasticity.

Abstract

In small-shot, replay-free incremental learning scenarios, the scarcity of training samples can lead to weak model feature representation and insufficient causal K-nearest neighbor candidate feature pools, resulting in weight estimation bias and prototype drift, thereby exacerbating catastrophic forgetting. Existing causality-constrained methods have good stability, but they lack plasticity under small-shot conditions; ordinary data augmentation only expands the training samples and lacks causal representation constraints. To address this, this paper proposes a prototype compensation method that combines attention-guided sample augmentation with causal constraints. Attention-guided data augmentation (AGDA) is used to generate semantically invariant synthetic samples, expanding the causal K-nearest neighbor candidate pool to optimize weight estimation; combined with causal regularization to protect old class knowledge, supervised contrastive loss is used to optimize the feature space, and prototype drift is corrected through adversarial shift compensation. Experiments were conducted on 4-task and 8-task small-shot settings of Caltech 101 and Oxford Flowers 102. The results show that, compared to the contrastive baseline, this method improves average accuracy by 1.2%–3.3% and average incremental accuracy by 1.0%–3.0%, effectively reducing the forgetting rate, providing a feasible solution path for replay-free incremental learning with scarce samples.

Downloads

Download data is not yet available.

References

[1] Huang, J., Xu, L., Liu, Y., et al. (2026). Reducing catastrophic forgetting via gradient-based task organization in class-incremental learning. Neurocomputing, 706, 134964. https://doi.org/10.1016/j.neucom.2026.134964.

[2] Wang, J., Hu, M., Li, N., Al-Ali, A., & Suganthan, P. N. (2026). Randomized neural network with adaptive forward regularization for online task-free class incremental learning. Neural Networks, 203, 109115. https://doi.org/10.1016/j.neunet.2026.109115

[3] Yuan, Z., Sun, B., Wang, Z., et al. (2026). Causal-enhanced machine learning framework for long-term wind power probabilistic forecasting. Reliability Engineering and System Safety, 272, 112532. https://doi.org/10.1016/j.ress.2026.112532

[4] Gudapati, M., Borah, A., & S, K. R. (2025). Incremental approach: A workflow model for semi-supervised learning environment. Engineering Applications of Artificial Intelligence, 160, 111815. https://doi.org/10.1016/j.engappai.2025.111815

[5] Jiang, M., Fan, J., & Li, F. (2025). Enhancing few-shot class-incremental learning through prototype optimization. Applied Intelligence, 55(10), 745. https://doi.org/10.1007/s10489-025-06605-z

[6] Li, D., Nair, C. K. N., & Wang, K. I. K. (2025). ADDAEIL: Anomaly detection with drift-aware ensemble-based incremental learning. Algorithms, 18(6), 359. https://doi.org/10.3390/a18060359

[7] Yuan, J., Chen, H., Tian, S., et al. (2025). Prompt-based learning for few-shot class-incremental learning. Alexandria Engineering Journal, 120, 287–295. https://doi.org/10.1016/j.aej.2025.02.008

[8] Ciarelli, M. P., Oliveira, E., & Salles, T. O. E. (2013). Impact of the characteristics of data sets on incremental learning. Artificial Intelligence Research, 2(4), 63–74. https://doi.org/10.5430/air.v2n4p63

[9] Hua, X., & Ding, S. (2011). Incremental learning algorithm for support vector data description. Journal of Software, 6(7), 1166–1173. https://doi.org/10.4304/jsw.6.7.1166-1173

[10] Zhang, H. S., Qu, L. X., & Wang, Y. X. (2011). Incremental regressive learning algorithm of support vector machine and its application. Advanced Materials Research, 216, 301–306. https://doi.org/10.4028/www.scientific.net/amr.216.301

[11] Devi, P., & Robi, P. (2007). An ensemble-based incremental learning approach to data fusion. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 37(2), 437–450.

[12] Hayes, T. L., Kafle, K., Shrestha, R., et al. (2020). REMIND your neural network to prevent catastrophic forgetting. In Computer Vision – ECCV 2020 (pp. 466–483). Springer.

[13] Al-Daoud, E. (2006). Incremental learning of auto-association multilayer perceptrons network. International Arab Journal of Information Technology, 3(1), 16–19.

[14] Hu, X. T., Tang, K. H., Miao, C. Y., et al. (2021). Distilling causal effect of data in class-incremental learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 3957–3966). IEEE. https://doi.org/10.1109/CVPR46437.2021.00395

[15] Liu, M., Quan, T., & Yao, T. (2006). Tracking maneuvering target based on neural fuzzy network with incremental neural learning. Journal of Systems Engineering and Electronics, 17(2), 343–349. https://doi.org/10.1016/S1004-4132(06)60060-1

[16] Sun, J., Wang, J., & Xu, Y. (2004). Incremental learning with balanced update on receptive fields for multi-sensor data fusion. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 34(1), 659–665.

[17] Lee, J., & Cho, S. (2003). An learning algorithm to find the optimized network structure in an incremental model. Journal of Internet Computing and Services, 4(5), 69–76.

[18] Lange, S., & Grieser, G. (2002). On the power of incremental learning. Theoretical Computer Science, 288(2), 277–307. https://doi.org/10.1016/S0304-3975(01)00404-2

[19] Zhang, J., Liu, L., Silven, O., et al. (2025). Few-shot class-incremental learning for classification and object detection: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence. https://doi.org/10.1109/TPAMI.2025.3529038

[20] Farah, J., Khaled, J., & Ouajdi, K. (2025). Intrusion detection based on concept drift detection and online incremental learning. International Journal of Pervasive Computing and Communications, 21(1), 81–115. https://doi.org/10.1108/ijpcc-12-2023-0358

[21] Taie, W., ElGeneidy, K., Yacoub, A. A., et al. (2024). Addressing catastrophic forgetting in payload parameter identification using incremental ensemble learning. Frontiers in Robotics and AI, 11, 1470163. https://doi.org/10.3389/frobt.2024.1470163

[22] Zhu, K., Xu, J., Zhou, L., et al. (2024). Dmaf: Data-model anti-forgetting for federated incremental learning. Cluster Computing, 28(1), 30. https://doi.org/10.1007/s10586-024-04697-9

[23] Kirkpatrick, J., Pascanu, R., Rabinowitz, N., et al. (2017). Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Sciences, 114(13), 3521–3526. https://doi.org/10.1073/pnas.1611835114

[24] Li, Z., & Hoiem, D. (2018). Learning without forgetting. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(12), 2935–2947. https://doi.org/10.1109/TPAMI.2017.2773087

[25] Zhao, B., Xiao, X., Gan, G., et al. (2020). Maintaining discrimination and fairness in class incremental learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 13208–13217). IEEE. https://doi.org/10.1109/CVPR42600.2020.01322

[26] Goswami, D., Soutif-Cormerais, A., Liu, Y., et al. (2024). Resurrecting old classes with new data for exemplar-free continual learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 23456–23465). IEEE. https://doi.org/10.1109/CVPR52733.2024.02474

[27] Douillard, A., Cord, M., Ollion, C., Robert, T., & Valle, E. (2020). PODNet: Pooled outputs distillation for small-tasks incremental learning. In Computer Vision – ECCV 2020 (pp. 86–102). Springer.

[28] Wang, N., Wang, H., Yang, S., et al. (2024). Semi-supervised incremental domain generalization learning based on causal invariance. International Journal of Machine Learning and Cybernetics, 15(10), 4815–4828. https://doi.org/10.1007/s13042-024-02199-z

[29] Sun, W., Li, Q., Zhang, J., et al. (2023). Class incremental learning based on identically distributed parallel one-class classifiers. Neurocomputing, 556. https://doi.org/10.1016/j.neucom.2023.126579.