Learning a Unified Classifier Incrementally via Rebalancing
183
stars
44
forks
source link
where is mimic score? In the paper Equation(5), author mentioned mimic score as an transformation of distillation loss. However I can't find it in the code. #11
ok, I find the mimic score loss, this loss isn't used combined with less forgetloss and marginrank loss