Domain Adaptation Metric Learning

For example si et al.

Domain adaptation metric learning. Specifically we present joint metric and feature representation learning jmfl for unsupervised domain adaptation. Domain adaptation in distance metric learning but neither learns a representation space capable of separating the source and target domains. Metric learning methods learn a metric to measure the distribution divergence between the source domain and the target domain and then reduce the divergence by minimizing the metric. Domain adaptation metric learning abstract.

Unsupervised domain adaptation is a promising avenue to enhance the performance of deep neural networks on a target domain using labels only from a source domain. Deep credible metric learning for unsupervised domain adaptation person re identi cation guangyi chen 1 2 3 yuhao lu 5 jiwen lu and jie zhou 4 1 department of automation tsinghua university china 2 state key lab of intelligent technologies and systems china 3 beijing national research center for information science and technology china 4 tsinghua shenzhen international graduate school. 14 select bregman divergence as a metric to estimate data distribution discrepancy between different domains then minimize the metric to match distributions. On the other hand jmfl reveals the shared underlying factors between the two domains to learn a new feature representation.

However the two predominant methods domain discrepancy reduction learning and semi supervised learning are not readily applicable when source and target domains do not share a common label space. Several methods for domain adaptation have recently been proposed arnold et al 2008. We thus propose a novel metric learning assisted domain adaptation mla da method which employs a novel triplet. The state of the art metric learning algorithms cannot perform well for domain adaptation settings such as cross domain face recognition image annotation etc because labeled data in the source domain and unlabeled ones in the target domain are drawn from different but related distributions.

Metric learning in optimal transport for domain adaptation tanguy kerdoncuff remi emonet and marc sebban univ lyon ujm saint etienne cnrs institut d optique graduate school laboratoire hubert curien umr 5516 f 42023 saint etienne france ftanguy kerdoncuff remi emonet marc sebbang univ st etienne fr abstract. The state of the art metric learning algorithms cannot perform well for domain adaptation settings such as cross domain face recognition image annotation etc because labeled data in the source domain and unlabeled ones in the target domain are drawn from different but related distributions. Resembling cd2ma luo et al 2017 considers domain adaptation with disjoint label spaces but the problem is still cast as classification with an assumption. Many existing da methods assume that a low source risk together with the alignment of distributions of source and target means a low target risk.

We provide extensive experimental results on real world datasets to both demonstrate the effectiveness of metric learning for graph construction for domain adaptation and analyze the contributions of each component.