Hierarchical optimal transport for unsupervised domain adaptation

DOI10.1007/S10994-022-06231-7OpenAlexW4298124876MaRDI QIDQ6097101FDOQ6097101

Authors Mourad El Hamri, Younès Bennani, Issam Falih

Publication date 12 June 2023

Published in Machine Learning (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2112.02073

zbMATH Keywords

optimal transportunsupervised learningdomain adaptationlearning probability measures

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Abstract: In this paper, we propose a novel approach for unsupervised domain adaptation, that relates notions of optimal transport, learning probability measures and unsupervised learning. The proposed approach, HOT-DA, is based on a hierarchical formulation of optimal transport, that leverages beyond the geometrical information captured by the ground metric, richer structural information in the source and target domains. The additional information in the labeled source domain is formed instinctively by grouping samples into structures according to their class labels. While exploring hidden structures in the unlabeled target domain is reduced to the problem of learning probability measures through Wasserstein barycenter, which we prove to be equivalent to spectral clustering. Experiments on a toy dataset with controllable complexity and two challenging visual adaptation datasets show the superiority of the proposed approach over the state-of-the-art.

Cites work

• scientific article; zbMATH DE number 1391397 (Why is no real title available?)
• scientific article; zbMATH DE number 7370561 (Why is no real title available?)
• scientific article; zbMATH DE number 7415088 (Why is no real title available?)
• scientific article; zbMATH DE number 3340881 (Why is no real title available?)
• scientific article; zbMATH DE number 2235418 (Why is no real title available?)
• scientific article; zbMATH DE number 3099866 (Why is no real title available?)
• A nonparametric ensemble transform method for Bayesian inference
• Advances in domain adaptation theory
• Barycenters in the Wasserstein space
• Computational optimal transport. With applications to data sciences
• Domain-adversarial training of neural networks
• Gromov-Wasserstein distances and the metric approach to object matching
• Improving predictive inference under covariate shift by weighting the log-likelihood function
• On the geometry of metric measure spaces. I
• Optimal Transport
• Optimal transport for applied mathematicians. Calculus of variations, PDEs, and modeling
• Quantization and the method of<tex>k</tex>-means
• Regularized discrete optimal transport
• The Sinkhorn–Knopp Algorithm: Convergence and Applications

Cited in

(2)

• MapFlow: latent transition via normalizing flow for unsupervised domain adaptation
• NaCL: noise-robust cross-domain contrastive learning for unsupervised domain adaptation

This page was built for publication: Hierarchical optimal transport for unsupervised domain adaptation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6097101)