The reason for setting _lambda ascending ?

[huster-wgm]

I just run the code using default setting of weight of _lambda , and i observe very tiny difference between testing accuracy of source and target. In the other hand, fixed value of 1 shows better result.

result of ascending _lambda ( _lambda = (e+1)/EPOCHS)

result of descending _lambda ( _lambda = (EPOCHS-e)/EPOCHS)

result of fixed _lambda ( _lambda = 1)

[SSARCandy]

I set the lambda ascending is because I want to focus on classification task at the beginning. In the original paper, the lambda is set in such way that at the end of training the classication loss and CORAL loss are roughly the same.

I think my linear lambda maybe too naive, or use a fixed lambda as your suggestion. Thanks!