Long-tailed representation learning algorithm based on adaptive prototypes and semantic awareness

Abstract

Label scarcity and long-tailed distribution imbalance are significant challenges in industrial equipment monitoring. Currently, self-supervised learning methods are affected by sample quantity bias and semantic confusion under complex operating conditions, which limits their ability to represent sparse critical states. To address these issues, we propose a co-evolutionary prototypical contrastive learning (EPCL) framework. Through progressive learning from coarse-grained semantic discovery to fine-grained discriminative enhancement, this framework enables an in-depth analysis of the intrinsic structure of long-tailed data. Specifically, an adaptive prototype-based clustering algorithm based on optimal transport theory is introduced, thereby achieving unbiased representation learning through data-driven dynamic priors. Furthermore, a semantic-aware and hierarchical negative sample weighting scheme is designed to optimize discriminative boundaries while mitigating class imbalance by enforcing prototype consistency constraints and employing an adaptive weighting strategy. Extensive experiments were conducted on several public long-tailed visual benchmarks, including CIFAR10-LT, CIFAR100-LT, and ImageNet-100-LT, as well as the industrial fault diagnosis dataset. The results demonstrated that the EPCL achieved better performance than fifteen mainstream self-supervised methods (e.g., SimCLR and SwAV) in both linear evaluation and few-shot classification tasks. On the CIFAR100-LT dataset , the EPCL improved the tail-class accuracy by 4.56% compared to SimCLR. Ablation studies and visualization results verified the effectiveness and generalization ability of the framework. This work offers a promising insight and practical solution for representation learning from unlabeled long-tailed measurement data.

Keywords

long-tailed distribution / self-supervised learning / contrastive learning / semantic awareness / adaptive prototype clustering / representation learning

Acknowledgement

The work was supported by the National Natural Science Foundation of China (No.12401703) and the Fundamental Research Program of Shanxi Province (Nos . 202203021211088, 202403021221109, 2024030 21212256).

Declaration of conflicting interests

The authors have no conflict of interests related to this publication.

References

[1] CHEN Z, CHEN J, FENG Y, et al. Imbalance fault diagnosis under long-tailed distribution: challenges, solutions and prospects. Knowledge-Based Systems, 2022, 258: 110008.

[2] CHAWLA N V, BOWYER K W, HALL L O, et al. SMOTE: synthetic minority over-sampling technique. Journal of Artificial Intelligence Research, 2002, 16(1): 321-357.

[3] SÁEZ J A, KRAWCZYK B, WOŹNIAK M. Analyzing the oversampling of different classes and types of examples in multi-class imbalanced datasets. Pattern Recognition, 2016, 57: 164-178.

[4] LIN C A, TSAI C F, LIN W C. Towards hybrid over- and under-sampling combination methods for class imbalanced datasets: an experimental study. Artificial Intelligence Review, 2023, 56(2): 845-863.

[5] CUI Y, JIA M L, LIN T Y, et al. Class-balanced loss based on effective number of samples//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 15-20, 2019. LongBeach, CA, USA. New York: IEEE, 2019: 9260-9269.

[6] PARK S, LIM J, JEON Y, et al. Influence-balanced loss for imbalanced visual classification//2021 IEEE/CVF International Conference on Computer Vision, October 10-17, 2021. Montreal, QC, Canada. New York: IEEE, 2021: 715-724.

[7] ZHU J G, WANG Z, CHEN J J, et al. Balanced contrastive learning for long-tailed visual recognition//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-24, 2022, New Orleans, LA, USA. New York: IEEE, 2022: 6898-6907.

[8] CAO K D, WEI C L, GAIDON A, et al. Learning imbalanced datasets with label-distribution-aware margin loss//The 33rd Conference on Neural Information Processing Systems, December 8-14, 2019, Vancouver, Canada. New York: Curran Associates, 2019: 1565-1576.

[9] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection//2017 IEEE International Conference on Computer Vision, October 22-29, 2017, Venice, Italy. New York: IEEE, 2017: 2999-3007.

[10] LI J, TAN Z C, WAN J, et al. Nested collaborative learning for long-tailed visual recognition//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-24, 2022, New Orleans, LA, USA. New York: IEEE, 2022: 6939-6948.

[11] LI T H, CAO P, YUAN Y, et al. Targeted supervised contrastive learning for long-tailed recognition//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 18-24, 2022, New Orleans, LA, USA. New York: IEEE, 2022: 6908-6918.

[12] FU S, CHU H, HE X, et al. Meta-prototype decoupled training for long-tailed learning//The Asian Conference on Computer Vision, December 4-8, 2022, Macao, China. Cham: Springer, 2022: 569-585.

[13] YANG Y Z, XU Z. Rethinking the value of labels for improving class-imbalanced learning//The 34th International Conference on Neural Information Processing Systems, December 6-12, 2020, Vancouver, BC, Canada. New York: ACM, 2020: 19290-19301.

[14] CARON M, MISRA I, MAIRAL J, et al. Unsupervised learning of visual features by contrasting cluster assignments. 2020: arXiv: 2006.09882.

[15] VAN DEN OORD A, LI Y Z, VINYALS O. Representation learning with contrastive predictive coding. 2018: arXiv: 1807.03748.

[16] DOERSCH C, ZISSERMAN A. Multi-task self-supervised visual learning//2017 IEEE International Conference on Computer Vision, October 22-29, 2017, Venice, Italy. New York: IEEE, 2017: 2070-2079.

[17] HE K M, FAN H Q, WU Y X, et al. Momentum contrast for unsupervised visual representation learning//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 13-19, 2020. Seattle, WA, USA. New York: IEEE, 2020: 9729-9738.

[18] CHEN T, KORNBLITH S, NOROUZI M, et al. A simple framework for contrastive learning of visual representations. 2020: arXiv: 2002.05709.

[19] CHUANG C Y, ROBINSON J, LIN Y C, et al. Debiased contrastive learning. Advances in Neural Information Processing Systems, 2020, 33: 8765-8775.

[20] GIDARIS S, SINGH P, KOMODAKIS N. Unsupervised representation learning by predicting image rotations. 2018: arXiv: 1803.07728.

[21] GRILL J B, STRUB F, ALTCHÉ F, et al. Bootstrap your own latent a new approach to self-supervised learning//The 34th International Conference on Neural Information Processing Systems, December 6-12, 2020, Vancouver, BC, Canada. New York: ACM, 2020: 21271-21284.

[22] CHEN X L, HE K M. Exploring simple Siamese representation learning//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 20-25, 2021. Nashville, TN, USA. New York: IEEE, 2021: 15745-15753.

[23] WANG T Z, ISOLA P. Understanding contrastive representation learning through alignment and uniformity on the hypersphere//The 37th International Conference on Machine Learning, New York: ACM, 2020: 9929-9939.

[24] KANG B Y, XIE S N, ROHRBACH M, et al. Decoupling representation and classifier for long-tailed recognition. 2019: arXiv: 1910.09217.

[25] WEI X S, SUN X H, SHEN Y, et al. Delving deep into simplicity bias for long-tailed image recognition. International Journal of Computer Vision, 2025, 133(6): 3349-3366.

[26] LIN C S, CHEN M H, WANG Y F. Frequency-aware self-supervised long-tailed learning//2023 IEEE/CVF International Conference on Computer Vision Workshops, October 2-6, 2023, Paris, France. New York: IEEE, 2023: 963-972.

[27] ZHOU Z, YAO J, WANG Y F, et al. Contrastive learning with boosted memorization//The 39th International Conference on Machine Learning, July 17-23, 2022, Baltimore, MD, USA. New York: PMLR, 2022: 27367-27377.

[28] ZHENG S F, NAM J, BAUR S, et al. Joint image clustering and self-supervised representation learning through debiased contrastive loss//Medical Imaging 2025: Image Processing, February 16-21, 2025. DiegoSan, USA. Bellingham: SPIE, 2025: 39.

[29] ZHAO Q, WU Z W, ZHANG Z Q, et al. Long-tail augmented graph contrastive learning for recommendation//Machine Learning and Knowledge Discovery in Databases: Research Track. Cham: Springer Nature Switzerland, 2023: 387-403.

[30] XIA Z Y, JIAN M, LIU Z H, et al. Mitigating long-tail bias in recommendations via graph diffusion. Multimedia Systems, 2025, 31(6): 468.

[31] SUN F H, LI G F, HE J L, et al. Intelligent diagnosis of high-speed motors under data imbalance scenarios: self-supervised feature extraction and classification optimization. The International Journal of Advanced Manufacturing Technology, 2025, 140(5): 3265-3278.

[32] JIANG Z, CHEN T, MORTAZAVI B J, et al. Self-damaging contrastive learning//The 38th International Conference on Machine Learning, July 18-24, 2021, Virtual. New York: PMLR, 2021, 139: 4927-4939.

[33] KIM D J, KE T W, YU S X. Local pseudo-attributes for long-tailed recognition. Pattern Recognition Letters, 2023, 172: 51-57.

[34] SMITH W A, RANDALL R B. Rolling element bearing diagnostics using the Case Western Reserve University data: a benchmark study. Mechanical Systems and Signal Processing, 2015, 64: 100-131.

[35] LI M K, HU Z K, LU Y, et al. Feature fusion from head to tail for long-tailed visual recognition. Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(12): 13581-13589.

[36] ZENG W, XIAO Z Y. MinoritySalMix and adaptive semantic weight compensation for long-tailed classification. Image and Vision Computing, 2024, 152: 105307.

[37] PAN H L, GUO Y, YU M J, et al. Enhanced long-tailed recognition with contrastive CutMix augmentation. IEEE Transactions on Image Processing, 2024, 33: 4215-4230.

[38] ZENG W, LI M. Leveraging multi-strategy labels for long-tailed classification. Engineering Applications of Artificial Intelligence, 2026, 166: 113563.

[39] ZHONG J, MAO H, MO F, et al. Uniform bullseye-guided contrastive learning with time-frequency strong augmentation for long-tailed fault diagnosis. Mechanical Systems and Signal Processing, 2025, 230: 112639.