Sameer Agarwal

One idea. I think its worth refactoring nist.cc into a problem solving as well as timing benchmark. both for ceres and for tinysolver. It is also worth thinking about doing...

I am quite confused by this error. Are you able to run bundle_adjuster ? with suite sparse as your sparse linear algebra library?

@NikolausDemmel sorry for the delayed reply. Time has been a bit fragmented lately. You raise some very interesting questions that I will try and address one at a time. >ceres-solver/internal/ceres/trust_region_minimizer.cc...

cc: @joydeep-b

Can you share the logs from the test please? I suspect it is a tolerance issue.

@alexsmac want to take this on? 2.0 is coming :)

@alexsmac are we doing this for 2.0 or should we punt on this?

How do you know your jacobians are correct? can you post the code thats computing it?

That makes no sense. On Wed, May 11, 2022, 3:32 AM LGR ***@***.***> wrote: > Problem solved. > > It does cause by jacobian. > > Here is the solution....

can you share the output of Summary::FullReport please? or better still run it with -v=2 and share the output. @alexsmac may have some insights.