python Keras sparse_categorical_crossentropy vs categorical_crossentropy

Posted 2021-05-09

# e.g. there are 10 categories
# y : array([9, 0, 0, ..., 3, 0, 5], dtype=uint8)
model.compile(loss='sparse_categorical_crossentropy',
              optimizer='adam', 
              metrics=['accuracy'])

# above is same as below
# one hot encoding first
y = keras.utils.to_categorical(y, 10).shape
model.compile(loss='categorical_crossentropy',
             optimizer='adam',
             metrics=['accuracy'])