Re-computed estimated measurements

[MaximeJ]

Hi @Serrof,

@Serrof

This seems a bit counterintuitive to me, especially if I have initialized the optimizer with an evaluation. Shouldn’t it be able to use it for the convergence check as well?

True, maybe line 306 of Hipparchus SequentialGaussNewton:

// iterate until convergence is reached
Evaluation current = null;

Should be replaced by:
Evaluation current = oldEvaluation

To be discussed with @bcazabonne I’d say.

I still think it would be simpler for you to get the last estimations with your own Python version of a ResidualsHandler.java (3.8 KB)