Really useful library! I'm using GPs for some large scale problems and wanted to be able to use your code. For exact GPs, it is well known that Cholesky decomposition is a better alternative than the base matrix inversion. I think the NumPyro example, which the ExactGP class borrows from, also reflects this now. Just added this slight change.
Let me know if I missed something. Testing the new test_gp.py runs successfully on my end.
I'm not sure if other parts of the codebase would also benefit from this.
Hi,
Really useful library! I'm using GPs for some large scale problems and wanted to be able to use your code. For exact GPs, it is well known that Cholesky decomposition is a better alternative than the base matrix inversion. I think the NumPyro example, which the ExactGP class borrows from, also reflects this now. Just added this slight change.
Let me know if I missed something. Testing the new
test_gp.py
runs successfully on my end.I'm not sure if other parts of the codebase would also benefit from this.