Obtain GaussNewtonOptimizer termination reasons

[raf...@dotproduct3d.com]

Hello,

the GaussNewtonOptimizer can terminate for reasons other than convergence, for example an error increase from one iteration to the next.

Is it possible to obtain the reason for optimizer termination after optimize() returns?

Not seeing it in the GaussNewtonOptimizer or NonlinearOptimizer classes but perhaps I'm missing something.

Thanks!

[raf...@dotproduct3d.com]

Never mind, I found a solution:

The NonlinearOptimizerParams offer an iterationHook that can be used to track error increases for any nonlinear optimizer.
It's just a bit hidden in between all the methods and constructors :)

[Dellaert, Frank]

There are also various verbosity settings.

Take a look at https://borglab.github.io/gtsam/levenbergmarquardtoptimizer/

The verbosity flag could​ be better documented, there. 

And the iteration hook seems t be not documented at all.

Hint hint 😊

Best!

Frank

--
You received this message because you are subscribed to the Google Groups "gtsam users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gtsam-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/gtsam-users/d2101f93-f34d-4be8-b485-1f1c7628136dn%40googlegroups.com.