(ranged) Sigmoid layer in advanced activation function #5452
Description
- Are you willing to contribute it (Yes I am willing to):
In tfjs, there is a lot of activation functions as a layer, but what is weirdly missing is layers.sigmoid.
This day I was making a dot product matrix factorization recommender system and the output of the dot layer should be in between 0 and 1, for this layers.sigmoid is needed.
FastAI also offer ranged sigmoid which is very interesting, it squeeze the value into a range ]a,b[ instead of ]1,0[, this can be also used in my case when i need to predict ratings in a certain range (let's say ratings from 1 star to 5 stars).
I can contribute to add this feature.