ABSTRACT

Dtr = {(xtr1 , ytr1 ), ..., (xtrm, ytrm)} ⊆ X ×Y, where X and Y denote the domains of predictors X and target Y , respectively. The estimated f is expected to generalize well on the test set Dte = {(xte1 , yte1 ), ..., (xten , yten )} ⊆ X ×Y, where ytei are unknown. Traditionally, the training set and test set are assumed to follow the same distribution. However, in many real world problems, the training data and test data have different distributions, i.e., P trXY 6= P teXY ,1 and the goal is to find a learning machine that performs well on the test domain. This problem is known as domain adaptation in machine learning.