search

thuml / Xlearn

Transfer Learning Library
465 stars 155 forks source link

DAN，JAN(caffe version) cannot be reproduced the results as papers reported #14

Closed liangzimei closed 6 years ago

liangzimei commented 6 years ago

hi, i run the DAN and JAN (alexnet, caffe version) on the A->W task, here is my result: Alexnet: 60% DAN: 65% JAN: 69% while the results in the paper is: DAN: 68% JAN: 74%

is there anything wrong, thanks in advance……

mengchuangji commented 6 years ago

I reproduce the results in this paper exactly,the result is the same.

liangzimei commented 6 years ago

@kind-hearted-Tbag do you use the caffe version?

mengchuangji commented 6 years ago

Yes. Loss_weight has a big impact on the results. You should search for the optimal Loss_weight with the strategy in the article. e.g.$\lambda$ in {0.01, 0.02,0.05,0.1,0.2,0.5,1.0} on task A$\rightarrow$W.

liangzimei commented 6 years ago

@kind-hearted-Tbag yeah, you are right , the loss_weight has a very big impact. thanks……