Hi @ricardoV94
I have updated the distribution to work directlly in log, which makes sure there is no overflow. The results are still way off, the distribution is basically unconstrained. I tested with a manual Gaussian distribution and it is working fine.
I prepared a clean notebook with my code, that you can in principle run directly (I had to change the extension to .txt). It is attached. Please let me know if you see where I went wrong.
Custom_Dist.txt (544.0 KB)