【摘要】 传统的基于自编码器的无监督领域自适应方法大多依靠单一的自编码器模型，故无法学习得到领域间的全局特征表示.针对该问题，提出一种基于串行自编码器(serial autoencoder unsupervised domain adaptation, SAUDA)的无监督领域自适应方法，以串行方式连接两种不同类型的自编码器学习更丰富的全局特征表示.利用堆叠自编码器(stacked autoencoder, SAE)对源域和目标域的特征进行初步学习；进一步地，采用稀疏自编码器(stacked sparse autoencoder, SSAE)对堆叠自编码器所得特征学习结果进行二次特征学习，以得到领域间更好的全局特征表示.结果表明，与传统的神经网络方法相比，基于SAUDA的无监督领域自适应方法在实验数据集上具有更好的跨领域分类性能.更多还原

【Abstract】 Traditional unsupervised domain adaptation methods based on autoencoders mostly rely on a single autoencoder model, so it is impossible to capture the global feature representation across domains. To address this problem, this paper proposes an unsupervised domain adaptation method based on serial autoencoder unsupervised domain adaptation(SAUDA), which connects two types of autoencoders serially to learn a richer global feature representation. Specifically, stacked autoencoder(SAE) is initially employed to learn features from the source and target domains. Subsequently, stacked sparse autoencoder(SSAE) is utilized to perform feature learning on the results obtained from the SAE, aiming to improve the global feature representation across domains. Experimental results demonstrate that the proposed method has better cross-domain classification performance than traditional neural network method on experimental data sets.更多还原