About supplementary material A for DTA

[smallQ-Q]

Thank professors for doing a so great work which is published on ICCV~:)

In the material A article, you professor used the 9CONV+1FC backbone for SVHN-->MNIST.

Here the problem which puzzles me is that what is the feature extractor and classifier when I used the 9CONV+1FC backbone? I tried to use 9CONV as the feature extractor and 1FC as the classifier. But it has a poor performance.
So could professor kindly tell me how should I choose the feature extractor and classifier?

Thank you in advance.