-
**Describe the bug 🐞**
`A \ b` works when `A` is of dimension $n\times p$ and $b$ is of dimension $n$. However, for `LinearSolve`, the code
```julia
using LinearSolve
lin_prob = LinearProblem(A,…
-
# Actual handling of inverses
In main's branch of MPQP, inverses are (nearly) not handled.
What I mean is that every time you want to have the inverse of a gate, it computes it **every** time, whi…
-
A python function that enables to compute maps of spatial gradients of a cost function (global in time, no time series of gradient maps which would require to prostproc numerical adjoint model and add…
pag13 updated
1 month ago
-
Hello, I have run my code by using odeint successfully, however when I use the odeint_adjoint, it comes out the error:
`RuntimeError: element 0 of tensors does not require grad and does not have a gr…
-
Hello,
I have been using the latest devel branch of seisflows to set up an example. The real and initial models are smaller, closer to exploration-sized models. I generated these models using specf…
-
We are missing an the inplace version of batched adjoint/transpose. They are required to avoid gpu scalar indexing with `Base.copy` like `copy(batched_adjoint(CUDA.randn(3,5,2)))`. They can be impleme…
-
This package suffers from severe type instability from some LinearAlgebra function.
Firstly, it seems that `eigen` is always performed on a symmetric matrix (A * A'), you should use `eigen(Hermitian(…
-
By adding this test:
```python
def test_wavelet_adjoint(wave_impl, shape_setup, floating_dtype):
"""Verify that the adjoint of the wavelet operator is correct.
This is done by checki…
-
In the tests we only check that the fourier transform is unitary,, not that the adjoint is correct. In fact, it is of by rather large factors in many cases. See e.g. this example:
```python
>>> od…
-
In recent main builds adjoint sensitivity analysis seems to give incorrect values compared to forward sensitivity methods as a result of upstream changes. #233