ja********@gmail.com wrote:
to**************@uniroma1.it wrote: It might not have solutions or have multiple (2) solution, depending on
A,B,C.
Thanks for your help, please let me know if you have any observations
about my adaptation.
Your function is like a lot of Access problems. I.e., different
conditions lend themselves to different solution techniques. Let's
look at the function.
f(x) = e ^ (Ax) - Ax - C
When x is very negative, f(x) looks similar to -Ax. When x is very
positive, f(x) looks like e ^ (Ax). The inflection point is when x =
0. The function is fairly flat near the inflection point. So f(x)
would be a somewhat reasonable way, for example, to approximate a
variable delay to an exponential function. The positive point where
f(x) = 0 can be thought of as the amount of time a delayed exponential
needs to break through a threshold. The minimum value of f(x) is 1 -
C, so C must be >= 1 for f(x) = 0 even to have a real solution. The
problem is that unless 1 - C is a reasonable distance below the x-axis,
the flat region has the potential to send p(k) off to places that even
Double will choke on. Once C is reasonably large, f(x) is so smooth
that convergence will happen quickly with almost any method. Note that
for this case the Newton-Raphson method will converge quite rapidly.
So I would check convergence for values of C slightly larger than one
with different values of A to see if Newton-Raphson is adequate for the
range of values you expect A to take.
James A. Fortune
CD********@FortuneJames.com