Grzego / erlang-ann

Simple artificial neural network written in erlang.
MIT License
0 stars 2 forks source link

Negative results #1

Open nopmop opened 3 years ago

nopmop commented 3 years ago

Hello and thank you for sharing this interesting piece of code. If I test with a large number of perceptrons & neurons I often get negative results for predictions. What does it mean? Is that normal?

Grzego commented 3 years ago

Hi @nopmop, by negative results you mean that the network never trains or that there are negative numbers in predictions?

nopmop commented 3 years ago

@Grzego hi, I meant negative numbers in predictions

Edit: I am using your ANN to analyze different celestial configurations (astronomical & astrological) to verify a conjecture regarding autism, to help children. I can show you what I am doing if you are interested.

Grzego commented 3 years ago

@nopmop regarding negative numbers. Yes, it's absolutely possible for the model to output those as the network here has a Linear layer as the output (which basically means no activation is applied to it).