Dear Authors:
In section 4.2.1 of the paper it is mentioned that NCM as the initial class prototypes for novel classes and these prototypes are the mean values of the features. And in the ETFHeadInc.py, etf_proto is initialized as a copy of etf_vec and I didn't find where the method set_proto is called. I would like to know how you are calculating prototypes? And I also would like to know how prototypes are updated, since the model parameters have changed?
Dear Authors:
In section 4.2.1 of the paper it is mentioned that NCM as the initial class prototypes for novel classes and these prototypes are the mean values of the features. And in the
ETFHeadInc.py,etf_protois initialized as a copy ofetf_vecand I didn't find where the methodset_protois called. I would like to know how you are calculating prototypes? And I also would like to know how prototypes are updated, since the model parameters have changed?