留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

Fea2Lab:基于多标记学习的特征到标签生成模型

于晗宇 黄晋 朱佳

于晗宇, 黄晋, 朱佳. Fea2Lab:基于多标记学习的特征到标签生成模型[J]. 华南师范大学学报(自然科学版), 2020, 52(3): 111-119. doi: 10.6054/j.jscnun.2020052
引用本文: 于晗宇, 黄晋, 朱佳. Fea2Lab:基于多标记学习的特征到标签生成模型[J]. 华南师范大学学报(自然科学版), 2020, 52(3): 111-119. doi: 10.6054/j.jscnun.2020052
YU Hanyu, HUANG Jin, ZHU Jia. Fea2Lab: A Feature-to-Label Generation Model Based on Multi-Label Learning[J]. Journal of South China normal University (Natural Science Edition), 2020, 52(3): 111-119. doi: 10.6054/j.jscnun.2020052
Citation: YU Hanyu, HUANG Jin, ZHU Jia. Fea2Lab: A Feature-to-Label Generation Model Based on Multi-Label Learning[J]. Journal of South China normal University (Natural Science Edition), 2020, 52(3): 111-119. doi: 10.6054/j.jscnun.2020052

Fea2Lab:基于多标记学习的特征到标签生成模型

doi: 10.6054/j.jscnun.2020052
基金项目: 

广东省自然科学基金项目 2016A030313437

广东省前沿与关键技术创新专项资金项目 2016B030305004

详细信息
    通讯作者:

    黄晋,副研究员,Email:jinhuang@scnu.edu.cn

  • 中图分类号: TP181

Fea2Lab: A Feature-to-Label Generation Model Based on Multi-Label Learning

  • 摘要: 传统多标记学习方法通常只考虑和示例相关联的单个特征向量以及无差别地预测全体标签,从而忽视了与示例相似的其他示例及隐含的标签属性,造成输入空间特征信息较少、标签属性被忽略和对大标记空间预测效果差等问题.为解决以上问题,文章转化传统多标记学习任务为多标记学习的序列到序列任务,并由此提出新的多标记学习标签生成神经网络模型(Fea2Lab模型):通过交错的顺序排列示例和相似示例形成链式特征向量序列,来增加输入空间特征信息;通过挖掘标签属性来有差别地预测标签;通过在解码流程中使用全局标签信息,来缓解预测过程中出现的错误标签级联问题.在多个数据集上的实验结果和消融实验表明转化任务和Fea2Lab模型的合理性、可行性及有效性.
  • 图  1  Fea2Lab模型计算流程图

    Figure  1.  The flow chart of the Fea2Lab model calculation

    图  2  多标记学习任务转化过程

    Figure  2.  The process of the multi-label learning task transformation

    图  3  4种特征链类型及对Fea2Lab模型AP值影响

    Figure  3.  The type of feature chain and its influence on the AP value of the Fea2Lab model

    图  4  全局标签信息对Fea2Lab模型的影响

    Figure  4.  The influence of global label information on the Fea2Lab model

    表  1  3个多标记数据集

    Table  1.   Three multi-label learning data-sets

    数据集 采集领域 示例数/个 特征维数/维 标签数/个 标签基数 标签密度 组合数/种
    Scene[4] 图像 2 407 294 6 1.074 0.179 15
    Yeast[7] 生物 2 417 103 14 4.237 0.303 198
    Emotions[23] 音乐 593 72 6 1.869 0.311 27
    下载: 导出CSV

    表  2  6个算法在3个数据集上的实验结果

    Table  2.   The experimental results of 6 algorithms on 3 datasets

    评价指标 算法 数据集
    Scene Yeast Emotions
    HL ML-KNN 0.084±0.008 0.195±0.011 0.194±0.013
    BR 0.104±0.006 0.199±0.010 0.199±0.022
    CC 0.096±0.010 0.208±0.010 0.192±0.027
    BP-MLL 0.282±0.014 0.205±0.009 0.219±0.021
    Fea2Lab 0.092±0.009 0.193±0.014 0.189±0.026
    LIFT 0.077±0.009 0.193±0.010 0.188±0.021
    Cov ML-KNN 0.078±0.010 0.447±0.014 0.300±0.019
    BR 0.089±0.009 0.514±0.018 0.295±0.027
    CC 0.091±0.008 0.516±0.015 0.322±0.022
    BP-MLL 0.374±0.024 0.456±0.019 0.300±0.022
    Fea2Lab 0.077±0.012 0.447±0.024 0.294±0.040
    LIFT 0.065±0.007 0.452±0.015 0.281±0.022
    RL ML-KNN 0.076±0.012 0.166±0.015 0.163±0.022
    BR 0.089±0.011 0.200±0.013 0.156±0.034
    CC 0.135±0.013 0.285±0.022 0.233±0.040
    BP-MLL 0.434±0.026 0.171±0.015 0.173±0.020
    Fea2Lab 0.074±0.007 0.169±0.007 0.161±0.018
    LIFT 0.062±0.008 0.163±0.011 0.144±0.024
    AP ML-KNN 0.869±0.017 0.765±0.021 0.799±0.031
    BR 0.849±0.016 0.749±0.019 0.807±0.037
    CC 0.852±0.016 0.728±0.019 0.796±0.042
    BP-MLL 0.445±0.018 0.754±0.020 0.779±0.227
    Fea2Lab 0.872±0.022 0.771±0.016 0.813±0.031
    LIFT 0.886±0.014 0.770±0.017 0.848±0.021
    下载: 导出CSV
  • [1] MITCHELL T M. Machine learning[J].New York:McGraw-Hill, 1997. http://d.old.wanfangdata.com.cn/NSTLQK/NSTL_QKJJ0216575822/
    [2] TSOUMAKAS G, KATAKIS I. Multi-label classification:an overview[J]. International Journal of Data Warehousing and Mining, 2007, 3(3):1-13. http://d.old.wanfangdata.com.cn/NSTLQK/NSTL_QKJJ0227335633/
    [3] ZHANG M, ZHOU Z. A review on multi-label learning algorithms[J]. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(8):1819-1837. doi: 10.1109/TKDE.2013.39
    [4] BOUTELL M R, LUO J, SHEN X, et al. Learning multi-label scene classification[J]. Pattern Recognition, 2004, 37(9):1757-1771. doi: 10.1016/j.patcog.2004.03.009
    [5] FVRNKRANZ J, HVLLERMEIER E, MENCÍA E L, et al. Multilabel classification via calibrated label ranking[J]. Machine Learning, 2008, 73(2):133-153. http://d.old.wanfangdata.com.cn/NSTLQK/NSTL_QKJJ027595258/
    [6] TSOUMAKAS G, VLAHAVAS I. Random, k-Labelsets: an ensemble method for multilabel classification[C]//Proceedings of the 18th European Conference on Machine Learning. Berlin: Springer, 2007: 406-417.
    [7] ELISSEEFF A, WESTON J. A kernel method for multi-labelled classification[C]//Proceedings of the 14th International Conference on Neural Information Processing Systems: Natural and Synthetic. Vancouver: MIT Press, 2001: 681-687.
    [8] ZHANG M L, ZHOU Z H. ML-KNN:a lazy learning approach to multi-label learning[J]. Pattern Recognition, 2007, 40(7):2038-2048. doi: 10.1016/j.patcog.2006.12.019
    [9] ZHANG M L, ZHANG K. Multi-label learning by exploiting label dependency[C]//Proceedings of the 16th ACM SIGKDD International Conference on Knowledge Disco-very and Data Mining. New York: ACM, 2010: 999-1008.
    [10] BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to align and translate[C/OL]//Proceedings of the 3rd International Conference on Learning Representations, (2014-09-01)[2019-10-10]. https://arxiv.org/abs/1409.0473.
    [11] SUN X, WEI B, REN X, et al. Label embedding network: learning label representation for soft training of deep networks[J/OL]. Computing Research Repository, (2017-10-28)[2019-10-10]. https://arxiv.org/abs/1710.10393.
    [12] YANG P, SUN X, LI W, et al. SGM: Sequence generation model for multi-label classification[C]//Proceedings of the 27th International Conference on Computational Linguistics. New-Mexico: ACL, 2018: 3915-3926.
    [13] NAM J, MENCÍA E L, KIM H J, et al. Maximizing subset accuracy with recurrent neural networks in multi-label classification[C]//Advances in Neural Information Processing Systems. California: Curran Associates, 2017: 5413-5423.
    [14] HUANG Y, WANG W, WANG L, et al. Multi-task deep neural network for multi-label learning[C]//Proceedings of 2013 IEEE International Conference on Image Processing. Melbourne: IEEE, 2013: 2897-2900.
    [15] ZHANG M L, ZHOU Z H. Multilabel neural networks with applications to functional genomics and text categorization[J]. IEEE Transactions on Knowledge and Data Enginee-ring, 2006, 18(10):1338-1351. doi: 10.1109/TKDE.2006.162
    [16] CHO K, VAN MERRIËNBOER B, GULCEHRE C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Qatar: ACL, 2014: 1724-1734.
    [17] VAN DER HEIJDEN F, DUIN R P, DE RIDDER D, et al. Classification, parameter estimation and state estimation: an engineering approach using MATLAB[M]. New Jersey:John Wiley & Sons, 2005.
    [18] 刘宇, 熊有伦.基于有界k-d树的最近点搜索算法[J].华中科技大学学报(自然科学版), 2008, 36(7):73-76. doi: 10.3321/j.issn:1671-4512.2008.07.020

    LIU Y, XIONG Y L. Algorithm for searching nearest-neighbor based on the bounded k-d tree[J]. Journal of Huazhong University of Science and Technology(Nature Science Edition), 2008, 36(7):73-76. doi: 10.3321/j.issn:1671-4512.2008.07.020
    [19] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735
    [20] LUONG M T, PHAM H, MANNING C D. Effective approaches to attention-based neural machine translation[C]//Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing.Qatar: ACL, 2015: 1412-1421.
    [21] YU L, ZHANG W, WANG J, et al. SeqGAN: sequence generative adversarial nets with policy gradient[C]//Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence. California: AAAI Press, 2017: 2852-2858.
    [22] FREITAG M, AL-ONAIZAN Y. Beam search strategies for neural machine translation[C]//Proceedings of the First Workshop on Neural Machine Translation. Vancouver: Association for Computational Linguistics, 2017: 56-60.
    [23] TROHIDIS K. Multi-labelclassification of music into emotions[C]//Proceedings of the 9th International Confe-rence on Music Information Retrieval. Philadelphia: [s.n], 2008: 325-330.
    [24] READ J, PFAHRINGER B, HOLMES G, et al. Classifier chains for multi-label classification[J]. Machine Lear-ning, 2011, 85(3):333-359. http://d.old.wanfangdata.com.cn/OAPaper/oai_arXiv.org_1304.5862
    [25] ZHANG M L, WU L. Lift:multi-label learning with label-specific features[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 37(1):107-120. http://d.old.wanfangdata.com.cn/Periodical/zgxdyxzz201802017
    [26] KINGMA D P, BA J. Adam: a method for stochastic optimization[C/OL]//Proceedings of the 3rd International Conference on Learning Representations, (2014-12-22)[2019-10-10]. https://arxiv.org/abs/1412.6980.
    [27] SRIVASTAVA N, HINTON G, KRIZHEVSKY A, et al. Dropout:a simple way to prevent neural networks from overfitting[J]. The Journal of Machine Learning Research, 2014, 15(1):1929-1958. http://d.old.wanfangdata.com.cn/Periodical/kzyjc200606005
  • 加载中
图(4) / 表(2)
计量
  • 文章访问数:  779
  • HTML全文浏览量:  575
  • PDF下载量:  36
  • 被引次数: 0
出版历程
  • 收稿日期:  2019-11-21
  • 刊出日期:  2020-06-25

目录

    /

    返回文章
    返回