Use Log Loss as Loss Function

I am working through the Dogs vs Cats Redux and I am wondering if I can use the log loss function as the Keras loss function. Currently I am using binary crossetropy and the loss seems very reasonable, but when I submit to Kaggle I am getting a much worse result so I’m hoping to replace my binary crossentropy loss with the log loss function.

Have you searched here? https://keras.io/losses/

Log loss and (binary) cross-entropy are different names for the same thing.

2 Likes