Open fwilliams opened 2 years ago
Hi @fwilliams ,
All my apologies for the long time to answer, Covid-Christmas has been pretty disruptive here in Paris :-)
To be clear, are you trying to solve a “block-wise” linear system where each block “at location (i,j)
” is given by the expression above, evaluated on points x_i
and x’_j
?
If this is the case, I believe that the simplest solution could be to:
What do you think?
Best regards, Jean
Hi, I'm working with Gaussian Processes with derivative observations and am trying to invert a kernel of the following form:
It would be nice to use
KernelSolve
in this case to invert the linear system. The expressions for the gradient parts of the kernel are fairly easy to write out as symbolic formulae, but I don't see an obvious way to write a reduction expression that I can plug directly intoKernelSolve
Any ideas?Thanks in advance!