reference
[1] A kernel two sample test
[2] Optimal kernel choice for large-scale two-sample tests
[3] Deep domain confusion: maximizing for domain invariance
[4] Learning transferable feature with deep adaptation nets
[5] Deep transfer network：Unsupervised domain adaptation
[6] Adaptive visual category models to new domains
[7] Geodesic flow kernel for unsupervised domain adaptation
[8] Transfer sparse coding for robust image representation
