(42S) HP 42S/ DM42: Nelder Mead Optimization

[pinkman]

Yes it helps, vielen dank! (Thanks a lot)

I added a counter line 37 before the call of the optimisation procedure, it’s quite funny to see that 2140 iterations are needed for a .0001 precision, but “only” 2191 for a precision of 10^-9

About you criterion calculation: so you use the distances between the original points of the simplex and not their image through FU.
It must be faster, but I’m really wondering about the consequences of this method, especially for non derivable functions.

Another question: you don’t care about the domain of FU? Inserting for example a square root in FU should create random comportment with complex values, don’t you think?

Regards,
Thibault