当前位置: X-MOL 学术J. Big Data › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Unsupervised feature learning-based encoder and adversarial networks
Journal of Big Data ( IF 8.6 ) Pub Date : 2021-09-06 , DOI: 10.1186/s40537-021-00508-9
Endang Suryawati 1 , Hilman F. Pardede 1 , Vicky Zilvan 1 , Ade Ramdan 1 , Dikdik Krisnandi 1 , Ana Heryana 1 , R. Sandra Yuwana 1 , R. Budiarianto Suryo Kusumo 1 , Andria Arisal 1 , Ahmad Afif Supianto 1
Affiliation  

In this paper, we propose a novel deep learning-based feature learning architecture for object classification. Conventionally, deep learning methods are trained with supervised learning for object classification. But, this would require large amount of training data. Currently there are increasing trends to employ unsupervised learning for deep learning. By doing so, dependency on the availability of large training data could be reduced. One implementation of unsupervised deep learning is for feature learning where the network is designed to “learn” features automatically from data to obtain good representation that then could be used for classification. Autoencoder and generative adversarial networks (GAN) are examples of unsupervised deep learning methods. For GAN however, the trajectories of feature learning may go to unpredicted directions due to random initialization, making it unsuitable for feature learning. To overcome this, a hybrid of encoder and deep convolutional generative adversarial network (DCGAN) architectures, a variant of GAN, are proposed. Encoder is put on top of the Generator networks of GAN to avoid random initialisation. We called our method as EGAN. The output of EGAN is used as features for two deep convolutional neural networks (DCNNs): AlexNet and DenseNet. We evaluate the proposed methods on three types of dataset and the results indicate that better performances are achieved by our proposed method compared to using autoencoder and GAN.



中文翻译:

基于无监督特征学习的编码器和对抗网络

在本文中,我们提出了一种新的基于深度学习的特征学习架构,用于对象分类。传统上,深度学习方法是通过监督学习来训练对象分类的。但是,这将需要大量的训练数据。目前有越来越多的趋势采用无监督学习进行深度学习。通过这样做,可以减少对大量训练数据可用性的依赖。无监督深度学习的一种实现是用于特征学习,其中网络旨在从数据中自动“学习”特征以获得可用于分类的良好表示。自编码器和生成对抗网络 (GAN) 是无监督深度学习方法的示例。然而对于 GAN,由于随机初始化,特征学习的轨迹可能会朝着不可预测的方向发展,使其不适合特征学习。为了克服这个问题,提出了编码器和深度卷积生成对抗网络 (DCGAN) 架构的混合架构,这是 GAN 的一种变体。编码器置于 GAN 的生成器网络之上,以避免随机初始化。我们称我们的方法为 EGAN。EGAN 的输出用作两个深度卷积神经网络 (DCNN) 的特征:AlexNet 和 DenseNet。我们在三种类型的数据集上评估了所提出的方法,结果表明,与使用自动编码器和 GAN 相比,我们提出的方法实现了更好的性能。提出了编码器和深度卷积生成对抗网络 (DCGAN) 架构的混合体,这是 GAN 的一种变体。编码器置于 GAN 的生成器网络之上,以避免随机初始化。我们称我们的方法为 EGAN。EGAN 的输出用作两个深度卷积神经网络 (DCNN) 的特征:AlexNet 和 DenseNet。我们在三种类型的数据集上评估了所提出的方法,结果表明,与使用自动编码器和 GAN 相比,我们提出的方法实现了更好的性能。提出了编码器和深度卷积生成对抗网络 (DCGAN) 架构的混合体,这是 GAN 的一种变体。编码器置于 GAN 的生成器网络之上,以避免随机初始化。我们称我们的方法为 EGAN。EGAN 的输出用作两个深度卷积神经网络 (DCNN) 的特征:AlexNet 和 DenseNet。我们在三种类型的数据集上评估了所提出的方法,结果表明,与使用自动编码器和 GAN 相比,我们提出的方法实现了更好的性能。

更新日期:2021-09-07
down
wechat
bug