Abstract

In this paper, we propose a new approach for domain gener- alization by exploiting the low-rank structure from multiple latent source domains. Motivated by the recent work on exemplar-SVMs, we aim to train a set of exemplar classifiers with each classifier learnt by using only one positive training sample and all negative training samples. While positive samples may come from multiple latent domains, for the posi- tive samples within the same latent domain, their likelihoods from each exemplar classifier are expected to be similar to each other. Based on this assumption, we formulate a new optimization problem by introduc- ing the nuclear-norm based regularizer on the likelihood matrix to the ob jective function of exemplar-SVMs. We further extend Domain Adap- tation Machine (DAM) to learn an optimal target classifier for domain adaptation. The comprehensive experiments for ob ject recognition and action recognition demonstrate the effectiveness of our approach for do- main generalization and domain adaptation.