1
$\begingroup$

I wanted to write a recursion function to calculate the n-th convolution of a function f[x_]. I wrote the function

g[f[x_], y_, z_] := Convolve[f[x], y, x, z]

where

f[x_] := \[Lambda]*
  Exp[-\[Lambda]*x/L]/((1 - Exp[-\[Lambda]])*L) UnitStep[x]

But now I want to put an n into the g function

g[f[x_], y_, z_,n_]

wich does the same n times! With DO[...] I have the problem the there is no output. How can I do it? thanks in advanced

$\endgroup$
3
  • $\begingroup$ Are you looking for Nest? $\endgroup$ Apr 7, 2016 at 13:39
  • $\begingroup$ I do not know Nest, do you think is it the solution? I checked it but the parameter z should be changed in every step, how can we manage it? $\endgroup$
    – maniA
    Apr 7, 2016 at 13:56
  • $\begingroup$ Have you seen this? $\endgroup$ Apr 7, 2016 at 14:01

2 Answers 2

2
$\begingroup$

You could also use:

cm[f_, s_, t_, n_] := 
 InverseLaplaceTransform[LaplaceTransform[f, s, t]^(n + 1), t, s]

(s>0):

e.g.

TableForm[Table[{j, cm[Exp[-a s] UnitStep[s], s, t, j]}, {j, 1, 10}], 
 TableHeadings -> {None, {"n", "n-fold Convolution"}}]

enter image description here

$\endgroup$
5
  • $\begingroup$ I'm surprised InverseLaplaceTransform[] even works… :) $\endgroup$ Apr 20, 2016 at 13:02
  • $\begingroup$ @J.M. Actually so was I...I am sure it is brittle but for some straight forward situations may be enough :) $\endgroup$
    – ubpdqn
    Apr 20, 2016 at 13:04
  • $\begingroup$ @ubpdqn ok your solution is more elegant :) just one question: dose it work with every function, I mean e.g. with truncated exponential too? $\endgroup$
    – maniA
    Apr 20, 2016 at 15:31
  • $\begingroup$ @maniA I am not sure how robust it is. It may not work for your purposes. It appears to work with a number of functions. Apologies for delay: time zones :) $\endgroup$
    – ubpdqn
    Apr 21, 2016 at 4:45
  • $\begingroup$ @ubpdqn no matter thanks any way :) ok I tick your response as the answer, however be warned guys that it may not work for the general case ;) then use just my proposal $\endgroup$
    – maniA
    Apr 25, 2016 at 16:50
0
$\begingroup$

Thanks to all who helped or tried to help me here, finally I found a solution. My special function was the exponential distribution but one can apply it for any arbitrary function.

f[x_] = \[Lambda]*Exp[-\[Lambda]*x] UnitStep[x];
g[x_] = \[Lambda]*Exp[-\[Lambda]*x] UnitStep[x];
convi[n_] := {Do[
    g[x[i]] = Convolve[f[x[i - 1]], g[x[i - 1]], x[i - 1], x[i]], {i, 
     1, n}], g[x[n]]}[[2]]

and if you excuse for instance

convi[4]

you get the 4-fold convolution.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.