forum.alglib.net

Is it possible to add a logit function to calculate with a zero intercept? Currently the calculation appears to be something like:

1/exp(-v1 + -v2 .. + -vn)

I believe the zero intercept version change would be:

1/exp(0 + -v1 + -v2 .. -vn)

-----------------

Would it be useful to add pluggable neural network activation functions using FADBAD++ ?? The idea would be to decouple the NNet activation function from the fixed sigmoid functions currrently used and add flexibility with user-defined activation functions.

Alglib is a great library!! Keep up the good work.