2009/7/22 Maxime Boissonneault <[email protected]>:
> Hello,
> I am reading the documentation to do multidimensionnal minimization with
> GSL. I think there is a mistake on that page :
> http://www.gnu.org/software/gsl/manual/html_node/Multimin-Algorithms-without-Derivatives.html
>
> It says "They construct n vectors p_i from the starting vector x and the
> vector step_size as follows..." According to the following, it should be
> "They construct n+1 vectors p_i from the..."
>
> Can someone who understand the algorithm better confirm ?
>

To form an n-dimensional simplex you need n+1 vectors. The initial
vector is p_0. The algorithm constructs an "additional" n vectors. I
agree that there is an ambiguity in the documentation. Also, the
subscript to step_size (in the formulas) is a bit confusing. Usually
when writing subscripts, the vector names are just a single letter.
(we write x_i not zyx_i )

 The improved construction could be

Starting from the initial vector p_0, the algorithm constructs an
additional n vectors p_i (i=1 through n) using the step size vector  s
as follows

  p_0 = (x_0, x_1, ..., x_n)
  p_1 = (x_0 + s_0, x_1, ..., x_n)
  p_2 = (x_0, x_1 + s_1, ..., x_n)
 ... = ...
  p_n = (x_0, x_1, ..., x_n + s_n)

What do you guys think? If necessary, I can submit a patch for this.

raju


_______________________________________________
Help-gsl mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-gsl

Reply via email to