search

nathanhubens / Autoencoders

Implementation of simple autoencoders networks with Keras
MIT License
234 stars 89 forks source link

Inquiring about choice of activation function #2

Closed rakeshKM closed 3 years ago

rakeshKM commented 4 years ago

Hi Nathan, Just a quick question, is there any logic behind the choice of sigmoid being the activation function in lats layer of AE. ?

nathanhubens commented 3 years ago

Hi Nathan, Just a quick question, is there any logic behind the choice of sigmoid being the activation function in lats layer of AE. ?

Hi ! Yes the sigmoid ensures that your output values are between 0 and 1, as it should be because they are pixel values.