2.765

2022影响因子

(CJCR)

  • 中文核心
  • EI
  • 中国科技核心
  • Scopus
  • CSCD
  • 英国科学文摘

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

协作式生成对抗网络

张龙 赵杰煜 叶绪伦 董伟

张龙, 赵杰煜, 叶绪伦, 董伟. 协作式生成对抗网络. 自动化学报, 2018, 44(5): 804-810. doi: 10.16383/j.aas.2018.c170483
引用本文: 张龙, 赵杰煜, 叶绪伦, 董伟. 协作式生成对抗网络. 自动化学报, 2018, 44(5): 804-810. doi: 10.16383/j.aas.2018.c170483
ZHANG Long, ZHAO Jie-Yu, YE Xu-Lun, DONG Wei. Co-operative Generative Adversarial Nets. ACTA AUTOMATICA SINICA, 2018, 44(5): 804-810. doi: 10.16383/j.aas.2018.c170483
Citation: ZHANG Long, ZHAO Jie-Yu, YE Xu-Lun, DONG Wei. Co-operative Generative Adversarial Nets. ACTA AUTOMATICA SINICA, 2018, 44(5): 804-810. doi: 10.16383/j.aas.2018.c170483

协作式生成对抗网络

doi: 10.16383/j.aas.2018.c170483
基金项目: 

浙江省自然科学基金 LZ16F030001

浙江省国际合作项目 2013C24027

国家自然科学基金 61571247

详细信息
    作者简介:

    张龙  宁波大学博士研究生.2008年获得瑞典布京理工学院硕士学位.主要研究方向为神经网络与深度学习.E-mail:1401082013@nbu.edu.cn

    叶绪伦  宁波大学博士研究生.2016年获得宁波大学硕士学位.主要研究方向为非参聚类, 流形学习以及非负矩阵分解.E-mail:1601082017@nbu.edu.cn

    董伟  宁波大学硕士研究生.2015年获得辽宁科技大学学士学位.主要研究方向为神经网络, 深度学习.E-mail: 1511082629@nbu.edu.cn

    通讯作者:

    赵杰煜宁波大学教授.主要研究方向为计算机图像处理, 机器学习, 神经网络.本文通信作者.E-mail:zhao jieyu@nbu.edu.cn

Co-operative Generative Adversarial Nets

Funds: 

National Natural Science Foundation of Zhejiang Province LZ16F030001

International Cooperation Projects of Zhejiang Province 2013C24027

National Natural Science Foundation of China 61571247

More Information
    Author Bio:

     Ph. D. candidate at Ningbo University. He received his master degree from Blekinge Institute of Technology, Sweden in 2008. His research interest covers neural networks and deep learning

     Ph. D. candidate at Ningbo University. He received his master degree from Ningbo University in 2016. His research interest covers non-parametric clustering, manifold learning, and non-negative matrix factorization

     Master student at Ningbo University. He received his bachelor degree from University of Science and Technology Liaoning in 2015. His research interest covers neural networks and deep learning

    Corresponding author: ZHAO Jie-Yu  Professor at Ningbo University. His research interest covers computer image processing, machine learning, and neural net-works. Corresponding author of this paper
  • 摘要: 生成对抗网络(Generative adversarial nets,GANs)将生成模型与判别模型进行了巧妙结合,采用无监督的训练方式,通过相互对抗共同提高,其在学术界掀起了一股新的机器学习热潮.GANs的学习目标是可以完整拟合任意真实样本的数据分布,然而在实际当中,真实样本分布的复杂程度难以预计,容易发生模式坍塌(Mode collapse)等问题,从而导致结果冗余,模型不收敛等.为提高无监督条件下的GANs生成能力,减少或消除模式坍塌,本文提出一种全新的协作式生成网络结构,通过构建多个生成模型,引入协作机制,使得生成模型在训练过程中能够相互学习,共同进步,从而提高模型对真实数据的拟合能力,进一步提高生成质量.通过在三组不同类型的数据集上进行实验,分析对比结果后发现新模型在二维图像生成方面,特别是人脸图片,有着显著的效果,协作机制不仅可以加快模型收敛速度,提高训练效率,还能消除损失函数噪声,在三维模型生成方面也产生了一定的影响.通过调整模型参数,模式坍塌问题也得到了遏制.本文还设计了一种动态学习方法,动态调节模型的学习速率,有效减少了过大或过小的梯度惩罚.
    1)  本文责任编委 李力
  • 图  1  生成对抗网络中的模式坍塌问题((a)生成数据分布无法完好拟合真实数据分布; (b)模式坍塌导致生成数据冗余(重复图像过多))

    Fig.  1  Mode collapse problem in GANs ((a) synthetic data distribution cannot fit real data distribution in good shape; (b) mode collapse leads to synthetic data redundancy (too many duplicate images))

    图  2  网络结构图

    Fig.  2  Network structure

    图  3  本文提出的网络拟合过程

    Fig.  3  Fitting process for proposed networks

    图  4  MNIST手写体数据集训练结果(上层采用标准生成对抗网络, 下层采用协作式生成对抗网络)

    Fig.  4  Training results on MNIST handwritten digits dataset (upper layer implements standard GANs, lower layer implements co-operative GANs

    图  5  CelebA人脸数据集训练结果(左侧为深度卷积生成对抗网络, 右侧为协作式生成对抗网络, (a)迭代500次; (b)迭代1 000次; (c) $\sim$ (h)迭代1 $\sim$ 6回合)

    Fig.  5  Training results on CelebA human faces dataset (left side is trained by DCGAN, right side is trained by ours after, (a) 500 iterations; (b) 1 000 iterations; (c) $\sim$ (h) 1 $\sim$ 6 epochs)

    图  6  CelebA数据集生成结果对比

    Fig.  6  Comparison of synthetic data with CelebA dataset

    图  7  判别与生成模型的损失函数值变换情况

    Fig.  7  Loss value changes of discriminator and generator models

    图  8  协作式生成对抗网络在ModelNet40数据集的训练结果

    Fig.  8  Results by co-operative GANs on ModelNet40 dataset

  • [1] Hinton G E. To recognize shapes, first learn to generate images. Progress in Brain Research, 2007, 165:535-547 doi: 10.1016/S0079-6123(06)65034-6
    [2] Taylor G W, Hinton G E, Roweis S. Modeling human motion using binary latent variables. In: Proceedings of the 19th International Conference on Neural Information Processing Systems. Canada: MIT Press, 2006. 1345-1352
    [3] Taylor G W, Hinton G E. Factored conditional restricted Boltzmann machines for modeling motion style. In: Proceedings of the 26th Annual International Conference on Machine Learning. Montreal, Quebec, Canada: ACM, 2009. 1025-1032
    [4] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks. Science, 2006, 313(5786):504-507 doi: 10.1126/science.1127647
    [5] Mohamed A, Dahl G E, Hinton G. Acoustic modeling using deep belief networks. IEEE Transactions on Audio, Speech, and Language Processing, 2012, 20(1):14-22 doi: 10.1109/TASL.2011.2109382
    [6] Hinton G, Deng L, Yu D, Dahl G E, Mohamed A R, Jaitly N, et al. Deep neural networks for acoustic modeling in speech recognition:the shared views of four research groups. IEEE Signal Processing Magazine, 2012, 29(6):82-97 doi: 10.1109/MSP.2012.2205597
    [7] Liu Y, Zhou S S, Chen Q C. Discriminative deep belief networks for visual data classification. Pattern Recognition, 2011, 44(10-11):2287-2296 doi: 10.1016/j.patcog.2010.12.012
    [8] Le Cun Y, Boser B, Denker J S, Howard R E, Habbard W, Jackel L D, et al. Handwritten digit recognition with a back-propagation network. In: Proceedings of Advances in Neural Information Processing Systems. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc., 1990. 396-404
    [9] Goodfellow I J, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, et al. Generative adversarial nets. In: Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal, Canada: MIT Press, 2014. 2672-2680
    [10] Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks. In: Proceedings of the 4th International Conference on Learning Representations. Caribe Hilton, San Juan, Puerto Rico, 2016. 97-108
    [11] Xue T F, Wu J J, Bouman K L, Freeman W T. Visual dynamics: probabilistic future frame synthesis via cross convolutional networks. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 91-99
    [12] Denton E L, Chintala S, Szlam A, Fergus R. Deep generative image models using a Laplacian pyramid of adversarial networks. In: Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal, Canada: MIT Press, 2015. 1486-1494
    [13] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks. In: Proceedings of Advances in Neural Information Processing Systems. Lake Tahoe, Nevada, USA: Curran Associates, Inc., 2012. 1097-1105
    [14] Liu M Y, Tuzel O. Coupled generative adversarial networks. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 469-477
    [15] Mirza M, Osindero S. Conditional generative adversarial nets. arXiv: 1411. 1784, 2014
    [16] van den Oord A, Kalchbrenner N, Espeholt L, Kavukcuoglu K, Vinyals O, Graves A. Conditional image generation with PixelCNN decoders. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 4790-4798
    [17] Reed S, Akata Z, Mohan S, Tenka S, Schiele B, Lee H. Learning what and where to draw. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 217-225
    [18] Salimans T, Goodfellow I, Zaremba W, Cheung V, Radford A, Chen X, et al. Improved techniques for training GANs. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 2226-2234
    [19] Chen X, Chen X, Duan Y, Houthooft R, Schulman J, Sutskever I, et al. InfoGAN: interpretable representation learning by information maximizing generative adversarial nets. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 2172-2180
    [20] Odena A, Olah C, Shlens J. Conditional image synthesis with auxiliary classifier GANs. In: Proceedings of the 34th International Conference on Machine Learning. Sydney, Australia, 2017. 2642-2651
    [21] Ghosh A, Kulharia V, Namboodiri V, Torr P H S, Dokania P K. Multi-agent diverse generative adversarial networks. arXiv: 1704. 02906, 2017
    [22] Arjovsky M, Bottou L. Towards principled methods for training generative adversarial networks. arXiv: 1701. 04862, 2017
    [23] LeCun Y, Cortes C, Burges C J C. The MNIST database of handwritten digits[Online], available: http://yann.lecun.com/exdb/mnist, June 3, 2017.
    [24] LeCun Y, Bottou L, Bengio Y, Haffner P. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 1998, 86(11):2278-2324 doi: 10.1109/5.726791
    [25] Liu Z W, Luo P, Wang X G, Tang X O. Large-scale CelebFaces Attributes (CelebA) Dataset[Online], available: http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html, July 20, 2017.
    [26] Liu Z W, Luo P, Wang X G, Tang X O. Deep learning face attributes in the wild. In: Proceedings of the 2015 IEEE International Conference on Computer Vision. Santiago, Chile: IEEE, 2015. 3730-3738
    [27] Princeton ModelNet[Online], available: http://modelnet.cs.princeton.edu, August 13, 2017.
    [28] Wu Z R, Song S R, Khosla A, Yu F, Zhang L G, Tang X O, et al. 3D ShapeNets: a deep representation for volumetric shapes. In: Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston, USA: IEEE, 2015. 1912-1920
    [29] Wu J J, Zhang C K, Xue T F, Freeman B, Tenenbaum J. Learning a probabilistic latent space of object shapes via 3D generative-adversarial modeling. In: Proceedings of Advances in Neural Information Processing Systems. Barcelona, Spain: Curran Associates, Inc., 2016. 82-90
  • 加载中
图(8)
计量
  • 文章访问数:  2091
  • HTML全文浏览量:  404
  • PDF下载量:  1394
  • 被引次数: 0
出版历程
  • 收稿日期:  2017-09-07
  • 录用日期:  2018-02-07
  • 刊出日期:  2018-05-20

目录

    /

    返回文章
    返回