-
A typical optimization process is evaluating usually the same derivatives at different points. Right now, the derivative is calculated from scratch each time (40sec in my case, where it should be less…
-
I have a somewhat complicated `torch.nn.Module`, let's say for arguments sake its structure is a bit like this:
```python
import torch
CustomModule(torch.nn.Module):
def __init__(self):
…
-
https://arxiv.org/abs/1506.05254
-
Multinomial logistic regression is implemented in `statsmodels` as `statsmodels.discrete.discrete_model.MNLogit`. After training, this model allows observations consisting of multiple quantitative fe…
-
More as a feature request: do you think it might be possible to allow some form of regularisation in `mblogit` multinomial fits?
When one is dealing with sparse data, one now often encounters problem…
-
Hi developers,
After I upgraded functorch from `v0.1.1` to `0.2.0`, I noticed a 25% performance regression when calculating hessian, please check the following benchmark result and the attached ben…
-
Hi all,
I would like to use Jax to compute the diagonal elelments of a Hessian matrix, i.e second partial derivatives \partial y^2 / \partial x_j^2. What's the most efficient way to do this? I know…
-
What is the proper approach to get performant hessian vector products for a scalar function with a vector valued input?
-
Manifolds and Optim walk into a bar asking for `Spheres`. Manifolds says the `norm` is to `project!`
https://github.com/JuliaManifolds/Manifolds.jl/blob/4c6cb43b9fce3ca4b506a00a6783aed9fc06a10f/src/…
-
At the moment I don't know if there is wide user interest, but feel free to contribute here.
See initial discussion with @timholy in https://github.com/JuliaDiff/AbstractDifferentiation.jl/pull/134