Hi, is the PALM loss only about unsupervised version? The protos update seems to use label information, whereas Q is not using labels. Then, the MLE loss is using the unsupervised version. Is there any specific reason for this combination?
More specifically,
https://github.com/jeff024/PALM/blob/main/util/loss_functions.py/#L106 the neg part the assignment to each prototypes w is not considered.
Thanks.
Hi, is the PALM loss only about unsupervised version? The protos update seems to use label information, whereas Q is not using labels. Then, the MLE loss is using the unsupervised version. Is there any specific reason for this combination?
More specifically,
https://github.com/jeff024/PALM/blob/main/util/loss_functions.py/#L106 the neg part the assignment to each prototypes w is not considered.
Thanks.