《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (12): 3662-3667.DOI: 10.11772/j.issn.1001-9081.2022121822
所属专题: 人工智能
李文博1,2, 刘波1,2(), 陶玲玲1,2, 罗棻1,2, 张航1,2
收稿日期:
2022-12-06
修回日期:
2023-02-20
接受日期:
2023-02-27
发布日期:
2023-03-03
出版日期:
2023-12-10
通讯作者:
刘波
作者简介:
李文博(1998—),男,重庆人,硕士研究生,主要研究方向:深度聚类、无监督学习、计算机视觉基金资助:
Wenbo LI1,2, Bo LIU1,2(), Lingling TAO1,2, Fen LUO1,2, Hang ZHANG1,2
Received:
2022-12-06
Revised:
2023-02-20
Accepted:
2023-02-27
Online:
2023-03-03
Published:
2023-12-10
Contact:
Bo LIU
About author:
LI Wenbo, born in 1998, M. S. candidate. His research interests include deep clustering, unsupervised learning, computer vision.Supported by:
摘要:
针对深度谱聚类模型训练不稳定和泛化能力弱等问题,提出L1正则化的深度谱聚类算法(DSCLR)。首先,在深度谱聚类的目标函数中引入L1正则化,使深度神经网络模型生成的拉普拉斯矩阵的特征向量稀疏化,并提升模型的泛化能力;其次,通过利用参数化修正线性单元激活函数(PReLU)改进基于深度神经网络的谱聚类算法的网络结构,解决模型训练不稳定和欠拟合问题。在MNIST数据集上的实验结果表明,所提算法在聚类精度(CA)、归一化互信息(NMI)指数和调整兰德系数(ARI)这3个评价指标上,相较于深度谱聚类算法分别提升了11.85、7.75和17.19个百分点。此外,所提算法相较于深度嵌入聚类(DEC)和基于对偶自编码器网络的深度谱聚类(DSCDAN)等算法,在CA、NMI和ARI这3个评价指标上也有大幅提升。
中图分类号:
李文博, 刘波, 陶玲玲, 罗棻, 张航. L1正则化的深度谱聚类算法[J]. 计算机应用, 2023, 43(12): 3662-3667.
Wenbo LI, Bo LIU, Lingling TAO, Fen LUO, Hang ZHANG. Deep spectral clustering algorithm with L1 regularization[J]. Journal of Computer Applications, 2023, 43(12): 3662-3667.
模块 | 网络结构 | |
---|---|---|
全连接层维度 | 激活函数 | |
自编码器模块 | ReLU | |
ReLU | ||
— | ||
特征映射模块 | ReLU | |
ReLU | ||
ReLU | ||
ReLU | ||
PReLU |
表1 DSCLR的网络结构
Tab.1 Network architecture of DSCLR
模块 | 网络结构 | |
---|---|---|
全连接层维度 | 激活函数 | |
自编码器模块 | ReLU | |
ReLU | ||
— | ||
特征映射模块 | ReLU | |
ReLU | ||
ReLU | ||
ReLU | ||
PReLU |
数据集 | 样本数 | 类别数 | 数据维度 |
---|---|---|---|
MNIST | 70 000 | 10 | 28×28×1 |
USPS | 9 298 | 10 | 28×28×1 |
FASHION | 70 000 | 10 | 28×28×1 |
DIGITS | 28 000 | 10 | 28×28×1 |
COIL20 | 1 440 | 20 | 128×128×1 |
表2 数据集详情
Tab.2 Dataset details
数据集 | 样本数 | 类别数 | 数据维度 |
---|---|---|---|
MNIST | 70 000 | 10 | 28×28×1 |
USPS | 9 298 | 10 | 28×28×1 |
FASHION | 70 000 | 10 | 28×28×1 |
DIGITS | 28 000 | 10 | 28×28×1 |
COIL20 | 1 440 | 20 | 128×128×1 |
算法 | MNIST | DIGITS | USPS | COIL20 | FASHION | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | |
k-means | 50.14 | 54.67 | 38.10 | 49.44 | 58.67 | 38.83 | 56.61 | 61.39 | 46.33 | 76.16 | 62.85 | 50.91 | 51.46 | 48.3 | 34.98 |
AC | 71.14 | 69.48 | 60.63 | 66.44 | 63.65 | 52.37 | 61.69 | 57.30 | 44.74 | 80.95 | 68.06 | 57.46 | 52.59 | 51.31 | 34.78 |
DBSCAN | 23.56 | 20.55 | 4.22 | 24.34 | 19.70 | 4.34 | 8.53 | 20.58 | 1.78 | 3.17 | 8.33 | 0.13 | 25.11 | 20.95 | 2.47 |
SC | 66.71 | 60.23 | 49.32 | 72.45 | 72.09 | 58.57 | 67.79 | 58.94 | 51.89 | 79.11 | 62.15 | 56.92 | 57.85 | 51.73 | 38.51 |
DEC | 84.40 | 81.60 | 79.50 | 78.40 | 80.05 | 76.33 | 61.90 | 58.60 | 55.40 | 61.00 | 62.10 | 55.32 | 57.81 | 62.83 | 45.71 |
SpectralNet | 73.75 | 72.17 | 63.63 | 84.11 | 87.91 | 82.47 | 63.53 | 60.65 | 53.03 | 50.47 | 46.95 | 49.20 | 58.41 | 66.06 | 46.30 |
SCDE | 83.31 | 79.02 | 81.31 | 86.79 | 85.91 | 84.19 | 64.67 | 67.04 | 56.97 | 64.56 | 67.86 | 54.56 | 49.29 | 66.45 | 49.29 |
DSCDAN | 77.60 | 74.50 | 78.34 | 85.40 | 80.10 | 86.25 | 69.80 | 65.20 | 57.56 | 47.60 | 64.20 | 53.64 | 54.60 | 55.60 | 40.82 |
DSCLR | 85.60 | 79.92 | 80.82 | 90.97 | 88.50 | 87.13 | 65.97 | 61.39 | 53.16 | 69.44 | 70.26 | 61.25 | 65.88 | 66.82 | 46.72 |
表3 各算法在不同数据集上的CA、NMI和ARI值 (%)
Tab.3 CA,NMI and ARI values of different algorithms on different datasets
算法 | MNIST | DIGITS | USPS | COIL20 | FASHION | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | CA | NMI | ARI | |
k-means | 50.14 | 54.67 | 38.10 | 49.44 | 58.67 | 38.83 | 56.61 | 61.39 | 46.33 | 76.16 | 62.85 | 50.91 | 51.46 | 48.3 | 34.98 |
AC | 71.14 | 69.48 | 60.63 | 66.44 | 63.65 | 52.37 | 61.69 | 57.30 | 44.74 | 80.95 | 68.06 | 57.46 | 52.59 | 51.31 | 34.78 |
DBSCAN | 23.56 | 20.55 | 4.22 | 24.34 | 19.70 | 4.34 | 8.53 | 20.58 | 1.78 | 3.17 | 8.33 | 0.13 | 25.11 | 20.95 | 2.47 |
SC | 66.71 | 60.23 | 49.32 | 72.45 | 72.09 | 58.57 | 67.79 | 58.94 | 51.89 | 79.11 | 62.15 | 56.92 | 57.85 | 51.73 | 38.51 |
DEC | 84.40 | 81.60 | 79.50 | 78.40 | 80.05 | 76.33 | 61.90 | 58.60 | 55.40 | 61.00 | 62.10 | 55.32 | 57.81 | 62.83 | 45.71 |
SpectralNet | 73.75 | 72.17 | 63.63 | 84.11 | 87.91 | 82.47 | 63.53 | 60.65 | 53.03 | 50.47 | 46.95 | 49.20 | 58.41 | 66.06 | 46.30 |
SCDE | 83.31 | 79.02 | 81.31 | 86.79 | 85.91 | 84.19 | 64.67 | 67.04 | 56.97 | 64.56 | 67.86 | 54.56 | 49.29 | 66.45 | 49.29 |
DSCDAN | 77.60 | 74.50 | 78.34 | 85.40 | 80.10 | 86.25 | 69.80 | 65.20 | 57.56 | 47.60 | 64.20 | 53.64 | 54.60 | 55.60 | 40.82 |
DSCLR | 85.60 | 79.92 | 80.82 | 90.97 | 88.50 | 87.13 | 65.97 | 61.39 | 53.16 | 69.44 | 70.26 | 61.25 | 65.88 | 66.82 | 46.72 |
1 | GUO X, LIU X, ZHU E, et al. Adaptive self-paced deep clustering with data augmentation[J]. IEEE Transactions on Knowledge and Data Engineering, 2019, 32(9): 1680-1693. 10.1109/TKDE.2019.2911833 |
2 | JAIN A K, MURTY M N, FLYNN P J. Data clustering: a review[J]. ACM Computing Surveys (CSUR), 1999, 31(3): 264-323. 10.1145/331499.331504 |
3 | 刘海杨,孟令航,林仲航,等.基于轨迹点聚类的航路发现方法[J]. 计算机应用, 2022, 42(3): 890-894. |
LIU H Y, MENG L H, LIN Z H, et al. GU. Route discovery method based on trajectory point clustering[J]. Journal of Computer Applications, 2022, 42(3): 890-894. | |
4 | 祝承,赵晓琦,赵丽萍,等. 基于谱聚类半监督特征选择的功能磁共振成像数据分类[J]. 计算机应用, 2021, 41(8): 2288-2293. 10.11772/j.issn.1001-9081.2020101553 |
ZHU C, ZHAO X Q, ZHAO L P, et al. Classification of functional magnetic resonance imaging data based on semi-supervised feature selection by spectral clustering[J]. Journal of Computer Applications, 2021, 41(8): 2288-2293. 10.11772/j.issn.1001-9081.2020101553 | |
5 | 袁野, 谭晓阳. 复杂环境下的冰箱金属表面缺陷检测[J]. 计算机应用, 2021, 41(1): 270-274. |
YUAN Y, TAN X Y. Defect detection of refrigerator metal surface in complex environment[J]. Journal of Computer Applications, 2021, 41(1): 270-274. | |
6 | 曾祥银, 郑伯川, 刘丹. 基于深度卷积神经网络和聚类的左右轨道线检测[J]. 计算机应用, 2021, 41(8): 2324-2329. 10.11772/j.issn.1001-9081.2021030385 |
ZENG X Y, ZHENG B C, LIU D. Detection of left and right railway tracks based on deep convolutional neural network and clustering[J]. Journal of Computer Applications, 2021, 41(8): 2324-2329. 10.11772/j.issn.1001-9081.2021030385 | |
7 | KANUNGO T, MOUNT D M, NETANYAHU N S, et al. An efficient k-means clustering algorithm: analysis and implementation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(7): 881-892. 10.1109/tpami.2002.1017616 |
8 | VON L U. A tutorial on spectral clustering[J]. Statistics and Computing, 2007, 17(4): 395-416. 10.1007/s11222-007-9033-z |
9 | ACKERMANN M R, BLÖMER J, KUNTZE D, et al. Analysis of agglomerative clustering[J]. Algorithmica, 2014, 69(1): 184-215. 10.1007/s00453-012-9717-4 |
10 | ESTER M, KRIEGEL H P, SANDER J, et al. A density-based algorithm for discovering clusters in large spatial databases with noise [EB/OL]. [2022-10-20]. . 10.1007/3-540-48238-5_5 |
11 | XIE J, GIRSHICK R, FARHADI A. Unsupervised deep embedding for clustering analysis[C]// Proceedings of the 33th International Conference on Machine Learning. [S.l.]: PMLR, 2016: 478-487. |
12 | YANG B, FU X, SIDIROPOULOS N D, et al. Towards K-means-friendly spaces: simultaneous deep learning and clustering[C]// Proceedings of the 34th International Conference on Machine Learning. [S.l.]: PMLR, 2017: 3861-3870. |
13 | SHAHAM U, STANTON K, LI H, et al. SpectralNet: spectral clustering using deep neural networks [C]// Proceedings of the 6th International Conference on Learning Representations. Vancouver: OpenReview.net, 2018, 1050: 10. |
14 | YANG X, DENG C, ZHENG F, et al. Deep spectral clustering using dual autoencoder network[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 4066-4075. 10.1109/cvpr.2019.00419 |
15 | HE K, ZHANG X, REN S, et al. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification [C]// Proceedings of the 2015 IEEE/CVF International Conference on Computer Vision. Washington, DC: IEEE Computer Society, 2015: 1026-1034. 10.1109/iccv.2015.123 |
16 | 苏常保,龚世才.一种基于卷积神经网络的谱聚类算法[J].安徽大学学报(自然科学版),2022,46(5):20-26. |
SU C B, GONG S C. A spectral clustering algorithm based on convolutional neural network[J]. Journal of Anhui University(Natural Sciences), 2022,46(5):20-26. | |
17 | YE X, ZHAO J, CHEN Y, et al. Bayesian adversarial spectral clustering with unknown cluster number [J]. IEEE Transactions on Image Processing, 2020, 29: 8506-8518. 10.1109/tip.2020.3016491 |
18 | JI P, ZHANG T, LI H,et al.Deep subspace clustering networks[C]// Proceedings of the 30th Advances in Neural Information Processing Systems. Long Beach: Curran Associates Inc., 2017:24-33. |
19 | ZHOU L, WANG S, BAI X,et al.Iterative deep subspace clustering[C]// Proceedings of the 2018 Joint IAPR International Workshop Structural,Syntactic,and Statistical Pattern Recognition. Cham: Springer, 2018:42-51. 10.1007/978-3-319-97785-0_5 |
20 | HUANG S, OTA K, DONG M, et al. MultiSpectralNet: spectral clustering using deep neural network for multi-view data[J]. IEEE Transactions on Computational Social Systems, 2019, 6(4): 749-760. 10.1109/tcss.2019.2926450 |
21 | BIANCHIF M, GRATTAROLA D, ALIPPI C. Spectral clustering with graph neural networks for graph pooling [C]// Proceedings of the 37th International Conference on Machine Learning. [S.l.]: PMLR, 2020: 874-883. |
22 | TIBSHIRANI R. Regression shrinkage and selection via the lasso: a retrospective[J]. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2011, 73(3): 273-282. 10.1111/j.1467-9868.2011.00771.x |
23 | TIBSHIRANI R. Regression shrinkage and selection via the lasso[J]. Journal of the Royal Statistical Society: Series B (Methodological), 1996, 58(1): 267-288. 10.1111/j.2517-6161.1996.tb02080.x |
24 | 刘建伟,李双成,付捷,等.L1范数正则化SVM聚类算法[J].计算机工程,2012,38(12):185-187. 10.3969/j.issn.1000-3428.2012.12.055 |
LIU J W, LI S C, FU J, et al. L1-norm regularized SVM clustering algorithm[J]. Computer Engineering, 2012,38(12):185-187. 10.3969/j.issn.1000-3428.2012.12.055 | |
25 | KAWALE J, BOLEY D. Constrained spectral clustering using L1 regularization[C]// Proceedings of the 2013 SIAM International Conference on Data Mining. Philadelphia,PA: SIAM, 2013: 103-111. 10.1137/1.9781611972832.12 |
26 | LEMHADRI I, RUAN F, TIBSHIRANI R. LassoNet: neural networks with feature sparsity[C]// Proceedings of the 24th International Conference on Artificial Intelligence and Statistics. [S.l.]: PMLR, 2021: 10-18. |
27 | LIU K, WANG H. High-order co-clustering via strictly orthogonal and symmetric L1-norm nonnegative matrix tri-factorization[C]//Proceedings of the 27th International Joint Conference on Artificial Intelligence. Menlo Park, CA: AAAI Press, 2018: 2454-2460. 10.24963/ijcai.2018/340 |
28 | ZHAO M, LIU J. Robust clustering with sparse corruption via ℓ2, 1, ℓ1 norm constraint and Laplacian regularization[J]. Expert Systems with Applications, 2021, 186: 115704. 10.1016/j.eswa.2021.115704 |
29 | BAGIROV A M, MOHEBI E. An algorithm for clustering using L1-norm based on hyperbolic smoothing technique[J]. Computational Intelligence, 2016, 32(3): 439-457. 10.1111/coin.12062 |
30 | DENG L. The MNIST database of handwritten digit images for machine learning research [J]. IEEE Signal Processing Magazine, 2012, 29(6):141-142.. 10.1109/msp.2012.2211477 |
31 | HULL J J. A database for handwritten text recognition research[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1994, 16(5): 550-554. 10.1109/34.291440 |
32 | COHEN G, AFSHAR S, TAPSON J, et al. EMNIST: extending MNIST to handwritten letters[C]// Proceedings of the 13th International Joint Conference on Neural Networks. Piscataway: IEEE, 2017: 2921-2926. 10.1109/ijcnn.2017.7966217 |
33 | XIAO H, RASUL K, VOLLGRAF R. Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms [EB/OL]. [2022-10-20]. . |
34 | NENE S A. Columbia object image library [DB/OL]. [2022-10-20]. . |
35 | DUAN L, MA S, AGGARWAL C, et al. Improving spectral clustering with deep embedding, cluster estimation and metric learning[J]. Knowledge and Information Systems, 2021, 63(3): 675-694. 10.1007/s10115-020-01530-8 |
36 | CAI D, HE X, HAN J. Locally consistent concept factorization for document clustering[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 23(6): 902-913. 10.1109/tkde.2010.165 |
37 | VINH N, EPPS J, BAILEY J. Information theoretic measures for clusterings comparison: is a correction for chance necessary?[C]//Proceedings of the 26th Annual International Conference on Machine Learning. New York: ACM, 2009: 1073-1080. 10.1145/1553374.1553511 |
38 | VAN DER MAATEN L, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9(11): 2579-2605. |
[1] | 李顺勇, 李师毅, 胥瑞, 赵兴旺. 基于自注意力融合的不完整多视图聚类算法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2696-2703. |
[2] | 贾洁茹, 杨建超, 张硕蕊, 闫涛, 陈斌. 基于自蒸馏视觉Transformer的无监督行人重识别[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2893-2902. |
[3] | 潘烨新, 杨哲. 基于多级特征双向融合的小目标检测优化模型[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2871-2877. |
[4] | 秦璟, 秦志光, 李发礼, 彭悦恒. 基于概率稀疏自注意力神经网络的重性抑郁疾患诊断[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2970-2974. |
[5] | 王熙源, 张战成, 徐少康, 张宝成, 罗晓清, 胡伏原. 面向手术导航3D/2D配准的无监督跨域迁移网络[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2911-2918. |
[6] | 黄云川, 江永全, 黄骏涛, 杨燕. 基于元图同构网络的分子毒性预测[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2964-2969. |
[7] | 刘禹含, 吉根林, 张红苹. 基于骨架图与混合注意力的视频行人异常检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2551-2557. |
[8] | 顾焰杰, 张英俊, 刘晓倩, 周围, 孙威. 基于时空多图融合的交通流量预测[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2618-2625. |
[9] | 石乾宏, 杨燕, 江永全, 欧阳小草, 范武波, 陈强, 姜涛, 李媛. 面向空气质量预测的多粒度突变拟合网络[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2643-2650. |
[10] | 赵亦群, 张志禹, 董雪. 基于密集残差物理信息神经网络的各向异性旅行时计算方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2310-2318. |
[11] | 徐松, 张文博, 王一帆. 基于时空信息的轻量视频显著性目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2192-2199. |
[12] | 孙逊, 冯睿锋, 陈彦如. 基于深度与实例分割融合的单目3D目标检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2208-2215. |
[13] | 吴筝, 程志友, 汪真天, 汪传建, 王胜, 许辉. 基于深度学习的患者麻醉复苏过程中的头部运动幅度分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2258-2263. |
[14] | 李欢欢, 黄添强, 丁雪梅, 罗海峰, 黄丽清. 基于多尺度时空图卷积网络的交通出行需求预测[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2065-2072. |
[15] | 张郅, 李欣, 叶乃夫, 胡凯茜. 基于暗知识保护的模型窃取防御技术DKP[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2080-2086. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||