Re: maple and fsolve : not good, big bad bug.




plouffe schrieb:

Hello,

Have you ever tried to solve a trans. equation like cos(x) - x = 0 ?

for that one the real solution is neat and fast to compute :

BUT, others are kind of more difficult, at least for Maple.

For example : -3+cosh(x)-sinh(x)/tanh(x) = 0

the solution found will depend on the number of Digits,
unfortunately, the solution varies with the number of digits too.

If Digits = 128, then the solution found is
-99003.0000
when Digits = 64 , then the result is :
329999.0000000000000000000000000000000000000000000000000000000000...

in both cases it is false.

I have a whole plethora of equations like that,

does someone has a way to solve these equations ?

NOTE : mathematica is not any better.


Derive has no problem detecting the absence of solutions when instructed
to solve the equation -3+cosh(x)-sinh(x)/tanh(x) = 0 symbolically.

I suppose you instructed Maple to find a solution numerically, and
without specifying a search interval. I imagine the search will probe
the left-hand side at the requested precision, until the constant 3
drops below the least significant digit. For unknown reasons (we don't
know the search strategy), with Digits = 128 on Maple this seems to
happen for cosh(-99003.0000) = 1.430743797 10^42996, whence x =
-99003.0000 is - quite correctly - returned for a solution.

I haven't waited for Derive's numerical solver to terminate with this
equation, but I suspect that here too something will break sooner or
later.

Martin.
.



Relevant Pages

  • Re: Whats wrong with this simple formula?
    ... And if we fully rationalize this number, we get the same rational number. ... number, only they return all 17 digits, while Excel only returns 15 digits. ... Excel rounds this to 15 digits as you've seen. ... Here is the Maple code; ...
    (microsoft.public.excel.misc)
  • Re: maple and fsolve : not good, big bad bug.
    ... for that one the real solution is neat and fast to compute: ... BUT, others are kind of more difficult, at least for Maple. ... the solution varies with the number of digits too. ... Difficult to comment on that last part without seeing the Mathematica ...
    (sci.math.symbolic)
  • Re: calling the Maple kernel
    ... Here is something that puzzles me about the interface between Matlab ... The second answer is the exact one (to 16 digits at least). ... Maple functions? ...
    (comp.soft-sys.matlab)
  • Digits of Accuracy: Maple Toolbox for Matlab
    ... In Maple, it is quite easy to set the maximum number of digits.... ... I would like to also keep a large number of digits available when working in MATLAB and/or when working in Maple and having MATLAB performing a few calculations. ... For example the following is an example to use Maple and have Matlab peform a matrix inversion and send the result back to Maple..... ...
    (comp.soft-sys.matlab)
  • Plouffes inverter : New program, version of sept. 18 2004
    ... I made a new version of my inverter program. ... The text file is to be used with Maple, ... and up to 110 digits. ... This portion ...
    (sci.math)