Re: Computing distance matrix robustly

[Tom Holroyd]

On Sun, 2 Oct 2005, [utf-8] S?ren Hauberg wrote:
> Sorry about replying to my own post, but Tom Holroyd, contacted me
> off-list with a solution. Simply replace
> sX2 = sum(X.^2, 2);
> with
> sX2 = sumsq(X, 2);
> in the vectorized version, and then things work.

I'm actually surprised that it worked.  But perhaps (in addition 
to being faster) it is true that the FPU's registers are actually 
wider than a double?  In that case, since X.^2 has to make 
a temp array, it forces a loss of precision.  I wouldn't expect 
1e-7 of roundoff anyway, though.  Unless the arrays in question 
are floats?

Dr. Tom Holroyd
"A man of genius makes no mistakes. His errors are volitional and
are the portals of discovery." -- James Joyce



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------