生成对抗网络GAN综述
|
摘要
人能够理解事物运动的方式,因此对事物未来发展的预测比机器准。不过,作为一种新的深度神经网络系统,GAN(Generative Adversarial Network)生成的数据非常逼真,连人也无法辨别数据是真实的还是生成的。从某种意义上讲,GAN为指导人工智能系统完成复杂任务提供了一种全新的思路,让机器成为了一个专家。首先,讨论了GAN的基本模型和一些改进的GAN模型;然后,展示了GAN在超分辨图像生成、由文本描述生成图像、艺术风格图像生成和短视频生成方面的应用成果;最后,探讨了GAN在理论、架构和应用方面所面临的问题和其未来的研究方向。
Humans can understand the way of movement,so they can predict the future development of things more accurately than machines.But GAN(Generative Adversarial Network) is a new neural Network system,its data are very lifelike,even people can't identify whether the data are real or generated.In a sense,GAN provides a brand new thought for guiding the artificial intelligence system to accomplish complex tasks,and makes the machine a specialist.In this paper,first of all,the basic model and some improvements model of GAN were discussed.Then,some application achievements of GAN were shown,such as the images generated by the super resolution,by a text description,by the artistic style and short video generated.Finally,some problems of theory,architecture,and application in the future research were discussed
Humans can understand the way of movement,so they can predict the future development of things more accurately than machines.But GAN(Generative Adversarial Network) is a new neural Network system,its data are very lifelike,even people can't identify whether the data are real or generated.In a sense,GAN provides a brand new thought for guiding the artificial intelligence system to accomplish complex tasks,and makes the machine a specialist.In this paper,first of all,the basic model and some improvements model of GAN were discussed.Then,some application achievements of GAN were shown,such as the images generated by the super resolution,by a text description,by the artistic style and short video generated.Finally,some problems of theory,architecture,and application in the future research were discussed
引文
[1] GOODFELLOW I,BENGIO Y,COURVILLE A.Deep Learning[M].Cambridge,UK:MIT Press,2016:23-34.
[2] LIU Q,ZHAI J H,ZHANG Z Z,et al.A Survey on Deep Reinforcement Learning[J].Chinese Journal of Computers,2017,40(1):1-28.(in Chinese)刘全,翟建伟,章宗长,等.深度强化学习综述[J].计算机学报,2017,40(1):1-28.
[3] GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al.Gene- rative adversarial nets[C]//Proceedings of the 2014 Conference on Advances in Neural Information Processing Systems 27.Montreal,Canada:Curran Associates,2014:2672-2680.
[4] YU L T,ZHANG W N,WANG J,et al.SeqGAN:sequence gene- rative adversarial nets with policy gradient[J/OL].https://arxiv.org/abs/1609.05473.
[5] SHAKIR M,LAKSHMINARAYANAN B.Learning in Implicit Generative Models[J/OL].https://openreview.net/pdf?id=B16Jem9xe.
[6] CHENG C.Interpretation of the GAN and its progress in 2016[EB/OL].https://zhuan lan.zhihu.com/p/25000523 refer=dlclass.
[7] HU W W,TAN Y.Generating adversarial malware examples for black-box attacks based on GAN[J/OL].https:// openreview.net/pdf?id=7xes.
[8] MIRZA M,OSINDERO S.Conditional Generative Adversarial Nets[J].Computer Science,2014,27(8):2672-2680.
[9] DENTON E L,CHINTALA S,FERGUS R.Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks[C]//Advances in Neural Information Processing Systems.2015:1486-1494.
[10] RAVANBAKHSH S,LANUSSE F,MANDELBAUM R,et al.Enabling Dark Energy Science with Deep Generative Models of Galaxy Images[J/OL].https://arxiv.org/abs/1609.05796.
[11] SEBASTIAN N,CSEKE B,TOMIOKA R.f-GAN:Training Generative Neural Samplers using Variational Divergence Minimization[J/OL].https://arxiv.org/abs/1606.00709.
[12] ZHAO J B,MICHAEL M,YANN L C.Energy-based Generative Adversarial Network[J/OL].https://arxiv.org/abs/1609.03126.
[13] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas,NV,USA:IEEE,2016:770-778.
[14] CHEN X,DUAN Y,HOUTHOOFT R,et al.InfoGAN:Inter- pretable Representation Learning by Information Maximizing Generative Adversarial Nets[J/OL].https://arxiv.org/abs/1606.03657.
[15] GANIN Y,USTINOVA E,AJAKAN H,et al.Domain adversa- rial training of neural networks[J].Journal of Machine Learning Research,2016,17(59):1-35
[16] TOBIAS J.Unsupervised and semi-supervised learning with categorical generative adversarial networks[C]//ICLR-2016,Springenberg.2016:876-884.
[17] CHEN W Z,WANG H,LI Y Y et al.Synthesizing training images for boosting human 3D pose estimation[C]//Proceedings of the 2016 Fourth International Conference on 3D Vision (3DV).Stanford,CA,USA:IEEE,2016:479-488.
[18] PROBST M.Generative Adversarial Networks in Estimation of Distribution Algorithms for Combinatorial Optimization[J/OL].https://arxiv.org/abs/1509.09235.
[19] GREGOR K,DANIHELKA I,GRAVES A,et al.DRAW: A recurrent neural network for image generation[J/OL].https://arxiv.org/abs/ 1502.04623.
[20] RADFORD A,METZ L,CHINTALA S.Unsupervised representation learning with deep convolutional generative adversarial networks[J/OL].https://arxiv.org/abs/ 1511.06434.
[21] AUGUSTUS O,OLAH C,SHLENS J.Conditi- onal Image Synthesis With Auxiliary Classifier GANs[J/OL].https://arxiv.org/abs/1610.09585.
[22] HANOCK K,ZHANG B T.Generating Images Part by Part with Composite Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1607.05387.
[23] KURAKIN A,GOODFELLOW I,BENGIO S.Adversarial exam- ples in the physical world[J/OL].https://arxiv.org/abs/1607.02533.
[24] ANTONIA C,BHARATH A A.Task Specific Adversarial Cost Function[J/OL].[2017-01-17].http://Creswell.com/caa?arXiv:1609.08661.
[25] CHE T,LI Y,JACOB A P,et al.Mode Regularized Generative Adversarial Networks[C/OL].[2017-01-30].https://openreview.net/pdf?id=HJKkY35le.
[26] IM D J,MA H,KIM C D,et al.Generative Adversarial Paralleli- zation[C/OL].https://openreview.net/pdf?id=Sk8J83oee.
[27] METZ L,POOLE B,PFAU D,et al.Unrolled Generative Adversarial Networks[C/OL].https://openre view.net/pdf?id=BydrOIcle.
[28] WARDE-FARLEY D,BENGIO Y.Improving Generative Ad- versarial Networks with Denoising Feature Matching[C/OL].https://openreview.net/pdf?id=S1X7nhsxl.
[29] CHRISTIAN L.Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network[J/OL].https://ar-xiv.org/abs/1609.04802.
[30] ALEXEY D,BROX T.Generating images with perceptual similarity metrics based on deep networks[J/OL].https://arxiv.org/abs/1602.02644.
[31] REED S,AKATA Z,YAN X,et al.Generative adversarial text to image synthesis[C]//International Conference on International Conference on Machine Learning.JMLR.org,2016:1060-1069.
[32] LARSEN A B L,S?NDERBY S K,WINTHER O.Autoenco- ding beyond pixels using a learned similarity metric[J/OL].[2015-11-02].https://arxiv.org/abs/1512.09300.
[33] VONDRICK C,PIRSIAVASH H,TORRALBA A.Generating Videos with Scene Dynamics[C]//NIPS-2016.Stanford,CA:IEEE,2016:562-570.
[34] SPRINGENBERG J T.Unsupervised and Semi-supervised Lear- ning withCategorical Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1511.06390.
[35] LEON A.Gatys,Alexander S.Ecker,Matthias Bethge.A Neural Algorithm of Artistic Style[J/OL].https://arxiv.org/abs/1508.06576.
[36] ZHU H,LI Q M,LI D Q.Facial Multi-landmarks Localization Based on Single Convolution Neural Network[J].Computer Scien-ce,2018,45(4):273-279.(in Chinese)朱虹,李千目,李德强.基于单个卷积神经网络的面部多特征点定位[J].计算机科学,2018,45(4):273-279.
[37] REN J,HU X F,LI N.Transfer Prediction Learning Based on Hybrid of SDA and SVR[J].Computer Science,2018,45(1):281-286.(in Chinese)任俊,胡晓峰,李宁.基于SDA与SVR混合模型的迁移学习预测算法[J].计算机科学,2018,45(1):281-286.
[38] WANG K F,GOU C,DUAN Y J,et al.Generative Adversarial Networks:The State of the Art and Beyond[J].Acta Automatica Sinica,2017,43(3):321-333.(in Chinese)王坤峰,苟超,段艳杰,等.生成式对抗网络GAN的研究进展与展望[J].自动化学报,2017,43(3):321-333.
[39] ODENA A.Semi-SupervisedLearning with Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1508.06576.
[40] WANG X,GUPTA A.Generative Image Modeling using Style and Structure Adversarial Networks[J/OL].https://arxiv.org/abs/1603.05631.
[41] DENTON E L,CHINTALA S,FERGUS R.Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks[C]//Advances in Neural Information Processing Systems.2015:1486-1494.
[42] EDWARDS H,STORKEY A.Censoring Representations with an Adversary[J/OL].[2015-01-26].https://arxiv.org/abs/1511.05897.
[43] ZHOU Z H,FENG J.Deep Forest:Towards An Alternative to Deep Neural Networks[J/OL].https://arxiv.org/abs/1702.08835.第46卷第3期
[2] LIU Q,ZHAI J H,ZHANG Z Z,et al.A Survey on Deep Reinforcement Learning[J].Chinese Journal of Computers,2017,40(1):1-28.(in Chinese)刘全,翟建伟,章宗长,等.深度强化学习综述[J].计算机学报,2017,40(1):1-28.
[3] GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al.Gene- rative adversarial nets[C]//Proceedings of the 2014 Conference on Advances in Neural Information Processing Systems 27.Montreal,Canada:Curran Associates,2014:2672-2680.
[4] YU L T,ZHANG W N,WANG J,et al.SeqGAN:sequence gene- rative adversarial nets with policy gradient[J/OL].https://arxiv.org/abs/1609.05473.
[5] SHAKIR M,LAKSHMINARAYANAN B.Learning in Implicit Generative Models[J/OL].https://openreview.net/pdf?id=B16Jem9xe.
[6] CHENG C.Interpretation of the GAN and its progress in 2016[EB/OL].https://zhuan lan.zhihu.com/p/25000523 refer=dlclass.
[7] HU W W,TAN Y.Generating adversarial malware examples for black-box attacks based on GAN[J/OL].https:// openreview.net/pdf?id=7xes.
[8] MIRZA M,OSINDERO S.Conditional Generative Adversarial Nets[J].Computer Science,2014,27(8):2672-2680.
[9] DENTON E L,CHINTALA S,FERGUS R.Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks[C]//Advances in Neural Information Processing Systems.2015:1486-1494.
[10] RAVANBAKHSH S,LANUSSE F,MANDELBAUM R,et al.Enabling Dark Energy Science with Deep Generative Models of Galaxy Images[J/OL].https://arxiv.org/abs/1609.05796.
[11] SEBASTIAN N,CSEKE B,TOMIOKA R.f-GAN:Training Generative Neural Samplers using Variational Divergence Minimization[J/OL].https://arxiv.org/abs/1606.00709.
[12] ZHAO J B,MICHAEL M,YANN L C.Energy-based Generative Adversarial Network[J/OL].https://arxiv.org/abs/1609.03126.
[13] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas,NV,USA:IEEE,2016:770-778.
[14] CHEN X,DUAN Y,HOUTHOOFT R,et al.InfoGAN:Inter- pretable Representation Learning by Information Maximizing Generative Adversarial Nets[J/OL].https://arxiv.org/abs/1606.03657.
[15] GANIN Y,USTINOVA E,AJAKAN H,et al.Domain adversa- rial training of neural networks[J].Journal of Machine Learning Research,2016,17(59):1-35
[16] TOBIAS J.Unsupervised and semi-supervised learning with categorical generative adversarial networks[C]//ICLR-2016,Springenberg.2016:876-884.
[17] CHEN W Z,WANG H,LI Y Y et al.Synthesizing training images for boosting human 3D pose estimation[C]//Proceedings of the 2016 Fourth International Conference on 3D Vision (3DV).Stanford,CA,USA:IEEE,2016:479-488.
[18] PROBST M.Generative Adversarial Networks in Estimation of Distribution Algorithms for Combinatorial Optimization[J/OL].https://arxiv.org/abs/1509.09235.
[19] GREGOR K,DANIHELKA I,GRAVES A,et al.DRAW: A recurrent neural network for image generation[J/OL].https://arxiv.org/abs/ 1502.04623.
[20] RADFORD A,METZ L,CHINTALA S.Unsupervised representation learning with deep convolutional generative adversarial networks[J/OL].https://arxiv.org/abs/ 1511.06434.
[21] AUGUSTUS O,OLAH C,SHLENS J.Conditi- onal Image Synthesis With Auxiliary Classifier GANs[J/OL].https://arxiv.org/abs/1610.09585.
[22] HANOCK K,ZHANG B T.Generating Images Part by Part with Composite Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1607.05387.
[23] KURAKIN A,GOODFELLOW I,BENGIO S.Adversarial exam- ples in the physical world[J/OL].https://arxiv.org/abs/1607.02533.
[24] ANTONIA C,BHARATH A A.Task Specific Adversarial Cost Function[J/OL].[2017-01-17].http://Creswell.com/caa?arXiv:1609.08661.
[25] CHE T,LI Y,JACOB A P,et al.Mode Regularized Generative Adversarial Networks[C/OL].[2017-01-30].https://openreview.net/pdf?id=HJKkY35le.
[26] IM D J,MA H,KIM C D,et al.Generative Adversarial Paralleli- zation[C/OL].https://openreview.net/pdf?id=Sk8J83oee.
[27] METZ L,POOLE B,PFAU D,et al.Unrolled Generative Adversarial Networks[C/OL].https://openre view.net/pdf?id=BydrOIcle.
[28] WARDE-FARLEY D,BENGIO Y.Improving Generative Ad- versarial Networks with Denoising Feature Matching[C/OL].https://openreview.net/pdf?id=S1X7nhsxl.
[29] CHRISTIAN L.Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network[J/OL].https://ar-xiv.org/abs/1609.04802.
[30] ALEXEY D,BROX T.Generating images with perceptual similarity metrics based on deep networks[J/OL].https://arxiv.org/abs/1602.02644.
[31] REED S,AKATA Z,YAN X,et al.Generative adversarial text to image synthesis[C]//International Conference on International Conference on Machine Learning.JMLR.org,2016:1060-1069.
[32] LARSEN A B L,S?NDERBY S K,WINTHER O.Autoenco- ding beyond pixels using a learned similarity metric[J/OL].[2015-11-02].https://arxiv.org/abs/1512.09300.
[33] VONDRICK C,PIRSIAVASH H,TORRALBA A.Generating Videos with Scene Dynamics[C]//NIPS-2016.Stanford,CA:IEEE,2016:562-570.
[34] SPRINGENBERG J T.Unsupervised and Semi-supervised Lear- ning withCategorical Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1511.06390.
[35] LEON A.Gatys,Alexander S.Ecker,Matthias Bethge.A Neural Algorithm of Artistic Style[J/OL].https://arxiv.org/abs/1508.06576.
[36] ZHU H,LI Q M,LI D Q.Facial Multi-landmarks Localization Based on Single Convolution Neural Network[J].Computer Scien-ce,2018,45(4):273-279.(in Chinese)朱虹,李千目,李德强.基于单个卷积神经网络的面部多特征点定位[J].计算机科学,2018,45(4):273-279.
[37] REN J,HU X F,LI N.Transfer Prediction Learning Based on Hybrid of SDA and SVR[J].Computer Science,2018,45(1):281-286.(in Chinese)任俊,胡晓峰,李宁.基于SDA与SVR混合模型的迁移学习预测算法[J].计算机科学,2018,45(1):281-286.
[38] WANG K F,GOU C,DUAN Y J,et al.Generative Adversarial Networks:The State of the Art and Beyond[J].Acta Automatica Sinica,2017,43(3):321-333.(in Chinese)王坤峰,苟超,段艳杰,等.生成式对抗网络GAN的研究进展与展望[J].自动化学报,2017,43(3):321-333.
[39] ODENA A.Semi-SupervisedLearning with Generative Adversarial Networks[J/OL].https://arxiv.org/abs/1508.06576.
[40] WANG X,GUPTA A.Generative Image Modeling using Style and Structure Adversarial Networks[J/OL].https://arxiv.org/abs/1603.05631.
[41] DENTON E L,CHINTALA S,FERGUS R.Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks[C]//Advances in Neural Information Processing Systems.2015:1486-1494.
[42] EDWARDS H,STORKEY A.Censoring Representations with an Adversary[J/OL].[2015-01-26].https://arxiv.org/abs/1511.05897.
[43] ZHOU Z H,FENG J.Deep Forest:Towards An Alternative to Deep Neural Networks[J/OL].https://arxiv.org/abs/1702.08835.第46卷第3期
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |