2
$\begingroup$

I have an infinite sequence (see the graphic) which I want to interpolate with an analytic function. Polynomial interpolation fails due to Runge phenomenon.

What else can I do?

enter image description here

  • 4
    It depends *greatly* on your infinite sequence. What *is* the infinite sequence you have?2011-07-30
  • 0
    How would you use a polynomial to interpolate an infinite sequence? If you could settle for a mere $C^\infty$ function, you could use a convolution of your function with a bump function.2011-07-30
  • 6
    Whatever else you do, it seems you should start out with an ansatz $f(x)=-g(x)\cos\pi x$ and interpolate $g$. That is if you mean "analytic" in the technical sense of the word -- if you're just looking for an expression in closed form, you could use $a_n=(-1)^{n-1}b_n$ and then interpolate $b_n$.2011-07-30
  • 0
    What about a complex function? I could imagine, that the given points are on some conic with hyperbolic decreasing circumference. (so that a curve spirals around the x-axis using the z-axis as well). If you could provide some (more!) data I could play a bit...2011-07-30
  • 0
    With that imagination of a conic I meant the following. Express a sequence of complex values $z_k$ by your y-values (using a sequential index k=1,2,3,... ) as $z_k=abs(y_k)*\exp(i*\pi*x_k) $ and try to find a logarithmic or exponential interpretation for $abs(y_k) $ in terms of k.2011-07-30
  • 0
    Could it be, that each second x-coordinate has a shift of -1/6 ? if I assume this, then my guessed coordinates (absolute values) have possibly an exponential fit.2011-07-30
  • 0
    Could you convolve it with the sinc function?2016-03-21

1 Answers 1