Re: [Bug-gsl] bug in gsl-1.13 (multidimensional minimization)

[Brian Gough]

At Mon, 28 Sep 2009 14:35:33 +0200,
Roman Parys wrote:
> I have found a problem in  multidimensional minimization package of GSL. 
> The new method gsl_multimin_fminimizer_nmsimplex2
> should consume O(N) memory, however  the function 
> gsl_multimin_fminimizer_alloc() tries to allocate a matrix of (N+1,N) size.
> In function nmsimplex_alloc()  there is a line (file simplex2.c) :
> 
> state->x1 = gsl_matrix_alloc (n + 1, n);
> 
> As a result of this problem, I'm not able to minimize a function of 
> about 50000 variables (32 bit system)

Thanks for the email.  I can see there is a problem with the
documentation -- the O(N) should refer to CPU usage only.  I don't
think there is any way to reduce the memory usage of the simplex
algorithm itself below O(N^2).  I will correct the documentation to
make it clearer anyway.

-- 
Brian Gough

GNU Scientific Library -
http://www.gnu.org/software/gsl/