This repository provides a link for the python implementation of the unsupervised domain adaptation approach JDME. Visit this link https://github.com/xiaonajin/Joint-Distribution-Matching-Embedding to obtain the JDME code.
Briefly speaking, JDME makes use of a Grasmannian point to match the source and target joint distributions for domain adaptation. The joint distribution similarity is measured by the Maximum Mean Discrepancy (MMD) distance with the kernel being the product of a feature kernel and a label kernel.
The details of this unsupervised domain adaptation approach can be found in the following work:
@article{Jin2020Joint,
title = {Joint distribution matching embedding for unsupervised domain adaptation},
journal = {Neurocomputing},
volume = {412},
pages = {115-128},
year = {2020},
issn = {0925-2312},
doi = {https://doi.org/10.1016/j.neucom.2020.05.098},
author = {Xiaona Jin and Xiaowei Yang and Bo Fu and Sentao Chen}
}