Issue about cross entropy loss in tensorflow

WeitaoVan / L-GM-loss

Implementation of our accepted CVPR 2018 paper "Rethinking Feature Distribution for Loss Functions in Image Classification"

MIT License

172 stars 34 forks source link

Issue about cross entropy loss in tensorflow #12

Open deepmo24 opened 5 years ago

deepmo24 commented 5 years ago

Hi, The cross entropy loss you used in Tensforflow is 'sparse_softmax_cross_entropy_with_logits()'. however, this loss function will perform a 'softmax' on 'logits'. Since your lgm loss doesn't contain a 'softmax' operation, I wonder whether the loss function is correctly used here?

I am looking forward to hearing from you.