-
@llfung talked to me today about his work on adjoint-accelerated programmable inference for large PDEs, and what would be needed on Turing's part to support that. As I understand it (and I know very l…
-
I think adding the option to solve with ODE using adjoints rather than backprop+checkpointing could be an interesting addition to the toolbox.
https://papers.nips.cc/paper_files/paper/2021/file/ad…
-
I'm getting an error when I try to apply actions and adjoints. Could you please help me with this?
The followiing is a minimal example and the error.
```
from firedrake import *
from firedrake.__…
-
In testing PSyclone 3.0, we find that forward code that looks like
```
vector(:) = vector(:) + constant(:) * scalar
```
where `vector` and `scalar` are active no longer adjoints to
```
do …
-
**Describe the bug 🐞**
I'm getting
```julia
ERROR: LoadError: autojacvec choice SciMLSensitivity.TrackerVJP(false) is not supported by GaussAdjoint
```
When I don't specify `sensealg`.
I thin…
-
Nonlinear optimization algorithms that leverage just the gradient information (i.e. "first-order methods") can have trouble traversing through the cost function as the hessian becomes ill-conditioned.…
-
We need to get gradient information out of Antioch. This will involve essentially take the form, d(outputs)/d(parameters), e.g. derivative of the species concentration with respect to a diffusion coef…
-
-
Moving here the issue from GridapODEs.jl (gridap/GridapODEs.jl#28)
-
@fverdugo @oriolcg We want to create adjoints of `FEOperator`s. In this sense, I propose a method like
```
adj_op = Adjoint(op,uh)
```
where `op` is a `FEOperator`, `uh` is the solution. We have d…