Hello,
I am also trying to do the same but I just wanted to confirm again.
In my case I want to use weighted linear combination of two losses of classes
eg: loss1 = 0.8* SigmoidCrossEntropyLoss and loss2 = 0.2*Softmaxwithloss and I would like to optimise both loss together
Net_Loss = loss1 + loss2
So as you said before, if I just give weights for each loss, caffe will automatically does the need full for different types of loss too ??
thank you