Learning Disentangled Semantic Representation for Domain Adaptation

Abstract

Domain adaptation is an important but challenging task. Most of the existing domain adaptation methods struggle to extract the domain-invariant representation on the feature space with entangling domain information and semantic information. Different from previous efforts on the entangled feature space, we aim to extract the domain invariant semantic information in the latent disentangled semantic representation (DSR) of the data. In DSR, we assume the data generation process is controlled by two independent sets of variables, i.e., the semantic latent variables and the domain latent variables. Under the above assumption, we employ a variational auto-encoder to reconstruct the semantic latent variables and domain latent variables behind the data. We further devise a dual adversarial network to disentangle these two sets of reconstructed latent variables. The disentangled semantic latent variables are finally adapted across the domains. Experimental studies testify that our model yields state-of-the-art performance on several domain adaptation benchmark datasets.

References in corpus (6)

• Learning Transferable Features with Deep Adaptation Networks
• Unsupervised Domain Adaptation by Backpropagation
• Deep Domain Confusion: Maximizing for Domain Invariance
• NICE: Non-linear Independent Components Estimation
• Adversarial Discriminative Domain Adaptation
• Domain Alignment with Triplets

Cited by in corpus (5)

• Semi-Supervised Disentangled Framework for Transferable Named Entity Recognition
• Learning to Match Distributions for Domain Adaptation
• Disentanglement Then Reconstruction: Learning Compact Features for Unsupervised Domain Adaptation
• Bi-Directional Generation for Unsupervised Domain Adaptation
• Adversarial Dual Distinct Classifiers for Unsupervised Domain Adaptation