Abstract:In many practical transfer learning scenarios, the feature distribution is different across the source and target domains (i.e. non-i.i.d.). Maximum mean discrepancy (MMD), as a domain discrepancy metric, has achieved promising performance in unsupervised domain adaptation (DA). We argue that MMD-based DA methods ignore the data locality structure, which, to some extent, would cause the negative transfer effect. The locality plays an important role in minimizing the nonlinear local domain discrepancy underlying the marginal distributions. For better exploiting the domain locality, a novel local generative discrepancy metric (LGDM) based intermediate domain generation learning called Manifold Criterion guided Transfer Learning (MCTL) is proposed in this paper. The merits of the proposed MCTL are four-fold: 1) the concept of manifold criterion (MC) is first proposed as a measure validating the distribution matching across domains, and domain adaptation is achieved if the MC is satisfied; 2) the proposed MC can well guide the generation of the intermediate domain sharing similar distribution with the target domain, by minimizing the local domain discrepancy; 3) a global generative discrepancy metric (GGDM) is presented, such that both the global and local discrepancy can be effectively and positively reduced; 4) a simplified version of MCTL called MCTL-S is presented under a perfect domain generation assumption for more generic learning scenario. Experiments on a number of benchmark visual transfer tasks demonstrate the superiority of the proposed manifold criterion guided generative transfer method, by comparing with other state-of-the-art methods. The source code is available in <a class="link-external link-https" href="https://github.com/wangshanshanCQU/MCTL" rel="external noopener nofollow">this https URL</a>.

A Framework for Supervised Heterogeneous Transfer Learning Using Dynamic Distribution Adaptation and Manifold Regularization

A Framework for Supervised Heterogeneous Transfer Learning Using Dynamic Distribution Adaptation and Manifold Regularization

Transfer Learning with Dynamic Distribution Adaptation

Learning Transferable and Discriminative Features for Unsupervised Domain Adaptation

Flexible deep transfer learning by separate feature embeddings and manifold alignment

Adaptation Regularization: A General Framework for Transfer Learning

A General Framework for Transfer Sparse Subspace Learning

Subdomain Adaptation With Manifolds Discrepancy Alignment

Homogeneous Online Transfer Learning with Online Distribution Discrepancy Minimization

Discriminative Transfer Learning Via Local and Global Structure Preservation

Deep into the Domain Shift: Transfer Learning Through Dependence Regularization

Manifold transfer subspace learning based on double relaxed discriminative regression

Supervised heterogeneous transfer learning using random forests

A Recent Survey of Heterogeneous Transfer Learning

Manifold Transfer Learning via Discriminant Regression Analysis

Multi-metric Domain Adaptation for Unsupervised Transfer Learning

Deep Transfer Learning: Model Framework and Error Analysis

Manifold Criterion Guided Transfer Learning via Intermediate Domain Generation

Transfer Independently Together: A Generalized Framework for Domain Adaptation

Transfer Feature Learning with Joint Distribution Adaptation