Open mrgloom opened 6 years ago
In which cases sparse_categorical_crossentropy or binary_crossentropy_with_logits should be used? Both of them support N classes? Why default keras losses: binary_crossentropy or categorical_crossentropy not used?
sparse_categorical_crossentropy
binary_crossentropy_with_logits
binary_crossentropy
categorical_crossentropy
In which cases
sparse_categorical_crossentropyorbinary_crossentropy_with_logitsshould be used? Both of them support N classes? Why default keras losses:binary_crossentropyorcategorical_crossentropynot used?