Recently, there have been a number of studies related to domain adaptation. However, the motivating ideas behind these methods are different. To connect the existing work and hence to better understand the problem, in the following sections, we organize the existing work into several categories from our own viewpoint. First, in Section ![[*]](crossref.gif)
, we consider a line of work that is based on instance weighting. In Section , we look at some work that bears strong resemblance to semi-supervised learning. In Section , we review another line of work that is based on changing the representation of 
. Section reviews work using Bayesian priors, and Section reviews work related to multi-task learning. In Section , ensemble methods for domain adaptation are considered.
The categories are ordered in this way so that methods in Section , Section and Section are generally applicable to unsupervised domain adaptation problems, while methods in Section and Section can only handle supervised domain adaptation problems.