Aiming at the problem that most instance-basedtransfer learning methods are difficult to estimate the distribution parametersand having poor generalization ability

a regularized discriminant transferlearning algorithm is proposed. Based on the discriminant analysis and semi-supervisedlearning theory

the semi-supervised Gauss kernel discriminant analysis method is studied bykernel method and regularization method

and the reusable samples aretransferred by constructing the revised embedding space. On the one hand

screeningsamples in the mapping space can solve the difficulty of estimating theparameters of domain distribution; on the other hand

introducing pseudo-labeleddata and defining the distance function can avoid over-fitting problems. Theexperimental results on text and non-text datasets validate that the proposed algorithm can effectivelyimprove the accuracy and generalization ability of transferring.