octave-maintainers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: unvech function for possible inclusion in Octave


From: Michael Creel
Subject: Re: unvech function for possible inclusion in Octave
Date: Fri, 07 Jul 2006 15:18:36 +0200
User-agent: Thunderbird 1.5.0.4 (X11/20060615)

I verified that this works, and that it can be a good deal faster depending on the dimension. So I made the change. A new version is attached. Please add your name and email address to the authorship.

Here's a script for testing speed:
reps = 100;
timings = zeros(reps,1);
for i = 1:reps
        tic();
        dim = 20;
        a = rand(dim,dim);
        a = a'*a;
        b = vech(a);
        c = unvech(b);
        # norm(a-c)
        timings(i) = toc();
endfor
mean(timings)

Michael

Hall, Benjamin wrote:
Since we're discussing vectorization v. loops...I think we can replace

        x = zeros(g,g);

        # fill in the symmetric matrix
        k = 1;
        for i = 1:g
                for j = i:g
                        x(i,j) = v(k);
                        if (i != j) x(j,i) = v(k); endif
                        k = k + 1;
                endfor
        endfor

with

        ii = repmat( 1:g, [g 1] );
        idx= find( ii <= ii' );
        
        x(idx) = v
        x = x + x' - diag(diag(x))


This might be a case where the vectorization is obscuring what's happening
-- but the speedup is big enough even for moderately sized matrices that is
probably worth it.  Plus, you can leave the original for-loop code commented
out close by to help eliminate any confusion.
## Copyright (C) 2006  Michael Creel <address@hidden>
##
## This program is free software; you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 2 of the License, or
## (at your option) any later version.
##
## This program is distributed in the hope that it will be useful,
## but WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this program; if not, write to the Free Software
## Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

## -*- texinfo -*-
## @deftypefn {Function File} {} unvech (@var{v})
## Performs the reverse of "vech". Generates a symmetric matrix from the lower
## triangular elements, received as a vector @var{v}.
## @end deftypefn

# Note: this uses a double loop. A C version would be a lot faster for large 
matrices.

function x = unvech (v)

        if (nargin != 1)
                usage ("unvech (v)");
        endif

        if (! isvector(v))
                usage ("unvech (v)");
        endif

        # find out dimension of symmetric matrix
        p = length(v);
        g = -(1 - sqrt(1 + 8*p))/2;

        if (mod(g,1) != 0)
                error("unvech: the input vector does not generate a square 
matrix");
        endif

        x = zeros(g,g);

        # fill in the symmetric matrix, the obvious way
#       k = 1;
#       for i = 1:g
#               for j = i:g
#                       x(i,j) = v(k);
#                       if (i != j) x(j,i) = v(k); endif
#                       k = k + 1;
#               endfor
#       endfor

        # fill in the symmetric matrix, a more clever way
        ii = repmat( 1:g, [g 1] );
        idx= find( ii <= ii' );
        x(idx) = v;
        x = x + x' - diag(diag(x));
endfunction

reply via email to

[Prev in Thread] Current Thread [Next in Thread]