Replies: 2 comments 8 replies
-
I have no idea if this would work with surrogate gradients but would be a very interesting question to ask. @fzenke do you know of anyone who's tried that? |
Beta Was this translation helpful? Give feedback.
2 replies
-
@thesamovar is this something you think we should try? Would be happy to give this a go if you could point me where to start :-) |
Beta Was this translation helpful? Give feedback.
6 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I wonder if we can use the Hessian of the models to ask questions about "robustness" of different (classes of) solutions to different kinds of parameter perturbations. For example, we can ask what can we learn from the Hessians of different models about "stiff" and "sloppy" directions in parameters space, about the curvature of the loss functions at different solutions, etc.
Computing these Hessians should be very easy with tools like PyTorch and JAX, and for models with relatively small number of parameters we can even try to visualize them (which can be useful sometimes).
Would be happy to hear people's thoughts on this.
Beta Was this translation helpful? Give feedback.
All reactions