astrodeepnet / sbi_experiments

Simulation Based Inference experiments
MIT License
3 stars 3 forks source link

Add notebook showcasing the impact of the differentiability of the NN architecture #4

Closed EiffL closed 2 years ago

EiffL commented 2 years ago

This notebook shows that for a given neural network using relu, if we try to train its gradients it doesnt work, but if we make the nn C\infty, then everything works fine.