google
Add Cholesky decomposition
We should support the Cholesky decomposition per https://numpy.org/doc/stable/reference/generated/numpy.linalg.cholesky.html#numpy.linalg.cholesky, adding a pivot argument in the same way as qr and svd.
Hi @Dbhasin1,
The function and tests first need to be added to the various backends, which live in tensornetwork/backends/blah. The first is the AbstractBackend, a base class that just throws NotImplementedError. The others all wrap functions from other numerical packages, the idea being to provide a means to quickly switch between these.
Inside the backend files you will see many other functions, such as qr. Our QR differs from e.g. NumPy's by the presence of a pivot argument that matricizes the input. Cholesky should have a similar mechanism.
Once added to the backends, the function should also be added to the tn.Tensor interface, which lives in tensornetwork/tensor.py with functions defined in tensornetwork/linalg/blah
Hello there @alewis I'm interested in working on this issue has a PR been done? If not I would like to work on this.
Hello, please ignore this PR I made a mistake, the changes implemented for the Cholesky decomposition will be from a different PR.
Hello, I am sorry for the mess but #886 is the correct PR I was working on disregard #885 thank you.
@alewis @mganahl I have added the NotImplementedError function and it's respective test and made a PR for you to review.
@alewis Alright I have made the fix, just waiting on the testing. Also, I will start implementing the Cholesky decomposition, quick question, why does Jax not have it's own decompositions file? Oh, could you add me as an assignee for this issue? Thank you
Ok, I see, then I will start adding the function to that file and work my way from there, thank you!
Hello, Can I use this issue as my first contribution? I'm new so please can someone guide me?
@prat1999 already opened a PR on this (maybe you guys can work together on this)