mkocabas / focal-loss-keras

Focal Loss implementation in Keras
MIT License
328 stars 74 forks source link

optimize computational complexity #22

Open developer0hye opened 4 years ago

developer0hye commented 4 years ago

alpha should be multiplied to mean error value. I think this order can reduce the computational complexity caused by vectorization. But, I didn't do experiment for this. I want your feedback!