Domain Adaptation in Alternating Structure Optimization (ASO) Algorithm

Taozheng Zhang, Xiaojie Wang, and Hui Tong

Keywords

ASO, semisupervised learning, auxiliary problems(APs), target problems(TPs)

Abstract

Recently, a semi-supervised learning algorithm called ASO (Alternating Structure Optimization) has been proposed, which belongs to linear structural learning. It utilizes a number of auxiliary problems (APs) with unlabelled data and then extracts the common structural parameter of APs to improve the performances of the target problems (TPs). In reality, we may have plentiful unlabeled data from source domains, while we need to deal with material from target domain. The distributions of them are quite different. We will focus on this problem in the paper. Theoretical analysis and experimental results both prove that the performances are still satisfying even when unlabeled data (APs) come from a different domain. Namely, ASO algorithm is provided with the property of domain adaptation.

Important Links:



Go Back