-
The current API to compute/get a hessian in matscipy is the following function:
```python
def get_hessian(self,
atoms: ase.Atoms,
format: str = 'sparse',
…
-
The "core" of `x/optym` is the convention `def thing(fg: callable)`, where `fg` returns `(cost, grad)` based on the parameter vector `x`.
This is in a way restrictive, since gradient-less optimizer…
-
### 🐛 Describe the bug
Hi,
I am trying to compute Hessian-vector products (hvps) between individual layers of a neural net, with multiple vectors. To get hvps with multiple vectors, I am using vma…
-
[Understanding black-box predictions via influence functions](https://arxiv.org/abs/1703.04730)
How can we explain the predictions of a black-box model? In this paper, we use influence functions -- a…
-
## Feature
Abstraction for lazy evaluation of tensors that can make use of special matrix structure for linear algebra operations, similar to Tensorflow's / scipy's `LinearOperator`.
## Motivati…
-
Hello,
I've been playing around with `autograd` and I'm having a blast. However I'm having some difficulty with extracting the diagonal of the Hessian.
This is my current code:
```python
fro…
-
Hi,
I was using `torch.func` in pytorch 2.0 to compute the Hessian-vector product of a neural network.
I first used `torch.func.functional_call` to define a functional version of the neural netw…
-
I am working on a project which requires me to calculate the trace of the Hessian of standard ResNet architectures. To this end I am using the Hutchinson method, which requires me to form the Hessian …
-
Hi! I encountered this problem while working on a rather complicated optimization problem in 200+ dimensions: the result of a 'successful' optimization violated the constraint. I use IPNewton followin…
-
understanding the geometry of transforms better:
- [ ] behavior of tail
- [ ] convexity
- [ ] (needs more thought and discussion)