On 12/23/2009 4:49:04 AM, Tom_Gutman wrote:
>Optimization has always been flaky, and in general should not be used.
That's because have bugs, a more clean optimiz option could be a very useful tool.
>Use explicit symbolic evaluation where appropriate.
This don't work in the exposed sample, and whith no apparent reason.
>The absolute value function is not differentiable everywhere.
Abs is a continuous function at the entire real axis and differentiable for all reals with a very small hole at the origin.
>You get all kinds of problems trying to differentiate it.
Only a discrete jump at the origin.
>The symbolic processor is quite capable of returning results that use functions that are only defined to the symbolic processor, and don't work with the numeric processor.
But this is not the case in the exposed examples: copy the symbolic answers in green and paste as function definitions then you can find that there are not problems in those assignations, so question is what is the problem that make unusable symbolic evaluation.
>You need to check symbolic results for such usages. Here the symbolic processor defines, and uses, a two operand form of ab and signum. This form is quite invalid for the numeric processor.
No, they are good definitions for numerics, I can plot this functions as they are defined.
>BTW, what version of MC are you actually using?
Mathcad 14 M020.
>This section is for MC14 bugs, but you posted a MC11 format file.
Saved as mc11, developed under mc14.
>Also, where did the Abs function you are using come from?
Can't understand this question.
Regards. Alvaro.