基于信息融合的半监督有序分类框架

doi:10.19678/j.issn.1000-3428.0069814

摘要/Abstract

摘要： 有序分类属于分类的一种,其要求类标签存在自然顺序,在很多领域例如电影分级、年龄估计都得到了广泛的研究。目前,大部分有序分类方法假设所有样本都被标记。但由于数据的特殊性,在实践中往往难以收集大量的标记数据,影响有序分类的性能。针对以上问题,提出一种结合额外信息的半监督有序分类框架。首先,利用未标记样本的顺序关系生成额外的偏序信息,并将偏序信息构建为有向图网络;然后使用图神经网络(GNN)聚合邻居信息,丰富节点表示,同时捕捉节点间的顺序关系,利用学习到的表示恢复偏序信息间的全局排名;接着使用高斯混合加权的方法对数据特征根据全局排名进行加权,并使用聚类方法为全局排名赋予伪标签,从而将这些信息合并到有序信息中;最后,使用有监督学习的有序分类模型进行年龄估计。在FGNET、Adience、UTKFace 3个数据集上的实验结果表明,该框架使用较少的标记数据便能够取得可靠的性能,在平均绝对误差(MAE)、准确率(Accuracy) 2个指标上相较于半监督学习基线方法均有提升:MAE在3个数据集上分别降低了0.05、0.04、0.04,Accuracy在3个数据集上分别提高了4.8、4.5、3.5百分点。

关键词: 有序分类, 图神经网络, 特征加权, 信息融合, 半监督学习

Abstract: In ordinal classification, class labels have a natural order. It has been widely studied in various fields, such as movie ratings and age estimation. Most existing methods assume that all samples are labeled. However, the unique nature of data often makes the collection of extensive labeled data challenging, thereby affecting the performance of ordinal classification. This study proposes a semi-supervised ordinal classification framework that incorporates additional information. The framework starts by generating partial order information from the relationships among unlabeled samples and constructing a directed graph network. Then, it uses Graph Neural Network (GNN) to aggregate neighbor information, enrich node representations, and capture the order between nodes, thereby recovering global rankings from partial order information. Subsequently, the method applies a Gaussian mixture model for feature weighting according to global rankings and employs clustering to assign pseudo labels by integrating this information into ordered information. Finally, the framework uses supervised learning models for ordinal classification tasks such as age estimation. Experiments on the FGNET, Adience, and UTKFace datasets show that the framework achieves reliable performance with fewer labeled data. It performs better than semi-supervised learning baselines in terms of Mean Absolute Error (MAE) and Accuracy. Specifically, MAE decreases by 0.05, 0.04, and 0.04, and Accuracy increases by 4.8, 4.5, and 3.5 percentage points on the three datasets, respectively.

Key words: ordinal classification, Graph Neural Network (GNN), feature weighting, information fusion, semi-supervised learning

中图分类号:

TP18

冉烔宇, 汤梦姿, 解庆, 刘永坚. 基于信息融合的半监督有序分类框架[J]. 计算机工程, 2026, 52(2): 287-298.

RAN Tongyu, TANG Mengzi, XIE Qing, LIU Yongjian. Semi-Supervised Ordinal Classification Framework Based on Information Fusion[J]. Computer Engineering, 2026, 52(2): 287-298.

https://www.ecice06.com/CN/Y2026/V52/I2/287

参考文献

[1] ZHU F, CHEN X C, CHEN S, et al. Relative margin induced support vector ordinal regression[J]. Expert Systems with Applications, 2023, 231: 120766.
[2] CHU W, KEERTHI S S. Support vector ordinal regression[J]. Neural Computation, 2007, 19(3): 792-815.
[3] BELLMANN P, SCHWENKER F. Ordinal classification: working definition and detection of ordinal structures[J]. IEEE Access, 2020, 8: 164380-164391.
[4] 王雅辉, 钱宇华, 刘郭庆. 基于模糊优势互补互信息的有序决策树算法[J]. 计算机应用, 2021, 41(10): 2785. WANG Y H, QIAN Y H, LIU G Q. Ordered decision tree algorithm based on fuzzy dominance complementarity mutual information[J]. Journal of Computer Applications, 2021, 41(10): 2785. (in Chinese)
[5] ZHU H P, SHAN H M, ZHANG Y H, et al. Convolutional ordinal regression forest for image ordinal estimation[J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 33(8): 4084-4095.
[6] WANG J H, CHENG Y, CHEN J T, et al. Ord2Seq: regarding ordinal regression as label sequence prediction[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2023: 5865-5875.
[7] CRAMMER K, SINGER Y. PRanking with ranking[J]. Advances in Neural Information Processing Systems, 2001, 14: 641-647.
[8] SHEN L B, JOSHI A K. Ranking and reranking with perceptron[J]. Machine Learning, 2005, 60: 73-96.
[9] VARGAS V M, GUTIÉRREZ P A, HERVAS-MARTINEZ C. Cumulative link models for deep ordinal classification[J]. Neurocomputing, 2020, 401: 48-58.
[10] SHI W L, GU B, LI X, et al. Quadruply stochastic gradient method for large scale nonlinear semi-supervised ordinal regression AUC optimization[C]//Proceedings of the AAAI Conference on Artificial Intelligence.[S. l.]: AAAI Press, 2020: 5734-5741.
[11] TSUCHIYA T, CHAROENPHAKDEE N, SATO I, et al. Semisupervised ordinal regression based on empirical risk minimization[J]. Neural Computation, 2021, 33(12): 3361-3412.
[12] CHEN H Y, JIA Y Z, GE J M, et al. Incremental learning algorithm for large-scale semi-supervised ordinal regression[J]. Neural Networks, 2022, 149: 124-136.
[13] STEWART N, BROWN G D A, CHATER N. Absolute identification by relative judgment[J]. Psychological Review, 2005, 112(4): 881.
[14] SADER M, VERWAEREN J, PEREZ-FERNANDEZ R, et al. Integrating expert and novice evaluations for augmenting ordinal regression models[J]. Information Fusion, 2019, 51: 1-9.
[15] TANG M Z, PEREZ-FERNANDEZ R, BAETS B. A comparative study of machine learning methods for ordinal classification with absolute and relative information[J]. Knowledge-Based Systems, 2021, 230: 107358.
[16] LIU T Y. Learning to rank for information retrieval[J]. Foundations and Trends^® in Information Retrieval, 2009, 3(3): 225-331.
[17] FOGEL F, D’ASPREMONT A, VOJNOVIC M. Spectral ranking using seriation[J]. Journal of Machine Learning Research, 2016, 17(88): 1-45.
[18] 张会影, 圣文顺. 基于标记适应的人脸年龄识别优化算法[J]. 计算机工程, 2025, 51(1): 174-181. ZHANG H Y, SHENG W S. Improved algorithm for facial age recognition based on label adaptation[J]. Computer Engineering, 2025, 51(1): 174-181. (in Chinese)
[19] TANG M, PÉREZ-FERNÁNDEZ R, BAETS B. Combining absolute and relative information with frequency distributions for ordinal classification[C]//Proceedings of the 18th International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems. Lisbon, Portugal: Springer International Publishing, 2020: 594-602.
[20] TANG M, PÉREZ-FERNÁNDEZ R, BAETS B. Ordinal classification with a spectrum of information sources[J]. Expert Systems with Applications, 2022, 208: 118163.
[21] TANG M, PÉREZ-FERNÁNDEZ R, BAETS B. Fusing absolute and relative information for augmenting the method of nearest neighbors for ordinal classification[J]. Information Fusion, 2020, 56: 128-140.
[22] TANG M, PÉREZ-FERNÁNDEZ R, BAETS B. Distance metric learning for augmenting the method of nearest neighbors for ordinal classification with absolute and relative information[J]. Information Fusion, 2021, 65, 72-83.
[23] ARLEGI R, DIMITROV D. Fair elimination-type competitions[J]. European Journal of Operational Research, 2020, 287(2): 528-535.
[24] BRADLEY R A, TERRY M E. Rank analysis of incomplete block designs: I. the method of paired comparisons[J]. Biometrika, 1952, 39(3/4): 324-345.
[25] LUCE R D. Individual choice behavior: a theoretical analysis[M].[S. l.]: Courier Corporation, 2005.
[26] CARON F, DOUCET A. Efficient Bayesian inference for generalized Bradley[WT6,3.75BZ]—[WT6BZ]Terry models[J]. Journal of Computational and Graphical Statistics, 2012, 21(1): 174-196.
[27] KENETT D Y, PERC M, BOCCALETTI S. Networks of networks[WT6,3.75BZ]—[WT6BZ]an introduction[J]. Chaos, Solitons & Fractals, 2015, 80: 1-6.
[28] CUCURINGU M. Sync-rank: robust ranking, constrained ranking and rank aggregation via eigenvector and SDP synchronization[J]. IEEE Transactions on Network Science and Engineering, 2016, 3(1): 58-79.
[29] BACCO C, LARREMORE D B, MOORE C. A physical model for efficient ranking in networks[J]. Science Advances, 2018, 4(7): eaar8260.
[30] D’ASPREMONT A, CUCURINGU M, TYAGI H. Ranking and synchronization from pairwise measurements via SVD[J]. Journal of Machine Learning Research, 2021, 22(19): 1-63.
[31] RIGUTINI L, PAPINI T, MAGGINI M, et al. SortNet: learning to rank by a neural preference function[J]. IEEE Transactions on Neural Networks, 2011, 22(9): 1368-1380.
[32] KÖPPEL M, SEGNER A, WAGENER M, et al. Pairwise learning to rank by neural networks revisited: Reconstruction, theoretical analysis and practical performance[C]//Proceedings of European Conference on Machine Learning and Knowledge Discovery in Databases. Würzburg, Germany: Springer International Publishing, 2020: 237-252.
[33] BURGES C, SHAKED T, RENSHAW E, et al. Learning to rank using gradient descent[C]//Proceedings of the 22nd International Conference on Machine Learning. New York, USA: ACM Press, 2005: 89-96.
[34] MAURYA S K, LIU X, MURATA T. Graph neural networks for fast node ranking approximation[J]. ACM Transactions on Knowledge Discovery from Data, 2021, 15(5): 1-32.
[35] HE Y X, GAN Q, WIPF D, et al. GNNRank: learning global rankings from pairwise comparisons via directed graph neural networks[C]//Proceedings of International Conference on Machine Learning.[S. l.]: PMLR, 2022: 8581-8612.
[36] TONG Z, LIANG Y, SUN C, et al. Digraph inception convolutional networks[J]. Advances in Neural Information Processing Systems, 2020, 33: 17907-17918.
[37] WERRIJ P, KAPTEIN R. Clustering ordinal survey data in a highly structured ranking[C]//Proceedings of Dutch-Belgian Information Retrieval Workshop 2016. New York, USA: ACM, 2016: 1-5.
[38] ALDAHDOOH R T, ASHOUR W. DIMK-means—"distance-based initialization method for k-means clustering algorithm"[J]. International Journal of Intelligent Systems and Applications, 2013, 5(2): 41.
[39] SHIN N H, LEE S H, KIM C S. Moving window regression: a novel approach to ordinal regression[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press 2022: 18760-18769.
[40] SHI X T, CAO W Z, RASCHKA S. Deep neural networks for rank-consistent ordinal regression based on conditional probabilities[J]. Pattern Analysis and Applications, 2023, 26(3): 941-955.
[41] SHEN W, GUO Y L, WANG Y, et al. Deep regression forests for age estimation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2018: 2304-2313.
[42] HU Z J, YANG Z Y, HU X F, et al. SimPLE: similar pseudo label exploitation for semi-supervised classification[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D.C., USA: IEEE Press, 2021: 15099-15108.
[43] SOHN K, BERTHELOT D, CARLINI N, et al. FixMatch: Simplifying semi-supervised learning with consistency and confidence[J]. Advances in Neural Information Processing Systems, 2020, 33: 596-608.

选择文件类型/文献管理软件名称

选择包含的内容