Experiment related information

[vinodkkurmi]

Hi,

1. Could you please share the experiment details of table 3 in the paper( 31 classes--> 25 classes), which classes you used in the target domain and how did you choose them.
2. In the office dataset, we do not have a validation set, so which accuracy you consider as final accuracy (i.e. do you consider the highest accuracy in one run or last accuracy of given the epoch).
   I will be thankful if you could share this information. Thanks

[vinodkkurmi]

@zhuhan1236 @longmingsheng @caoyue10
Could you please share the information about your experiments in the paper?