search

keunwoochoi / music-auto_tagging-keras

Music auto-tagging models and trained weights in keras/theano
MIT License
616 stars 142 forks source link

issue related to cost function #13

Closed sushien closed 7 years ago

sushien commented 7 years ago

In your paper you have specified to use binary cross-entropy function. Why is that and what will happen on using categorical cross-entropy function?

keunwoochoi commented 7 years ago

Because it is a multi-class classification, which is reformed as 50 independent binary classifications. Your problem is single-class one, you can use categorical cross-entropy.