Timezone: »

Asymmetric Tri-training for Unsupervised Domain Adaptation
Kuniaki Saito · Yoshitaka Ushiku · Tatsuya Harada

Mon Aug 07 01:30 AM -- 05:00 AM (PDT) @ Gallery #131

It is important to apply models trained on a large number of labeled samples to different domains because collecting many labeled samples in various domains is expensive. To learn discriminative representations for the target domain, we assume that artificially labeling the target samples can result in a good representation. Tri-training leverages three classifiers equally to provide pseudo-labels to unlabeled samples; however, the method does not assume labeling samples generated from a different domain. In this paper, we propose the use of an \textit{asymmetric} tri-training method for unsupervised domain adaptation, where we assign pseudo-labels to unlabeled samples and train the neural networks as if they are true labels. In our work, we use three networks \textit{asymmetrically}, and by \textit{asymmetric}, we mean that two networks are used to label unlabeled target samples, and one network is trained by the pseudo-labeled samples to obtain target-discriminative representations. Our proposed method was shown to achieve a state-of-the-art performance on the benchmark digit recognition datasets for domain adaptation.

Author Information

Kuniaki Saito (The University of Tokyo)
Yoshitaka Ushiku (The University of Tokyo)

Dr. Ushiku received his Ph.D. in Information Science and Technology from the University of Tokyo, under supervision of Tatsuya Harada. He then accepted the position of a research scientist at Nippon Telegraph and Telephone (NTT) Corporation, where he began a research on image understanding. After two years at NTT, he has been a lecturer at the University of Tokyo since 2016. Currently, he is working extensively on researches about integrations of vision and language.

Tatsuya Harada (The Univ. of Tokyo / RIKEN)

Related Events (a corresponding poster, oral, or spotlight)