help-octave
[Top][All Lists]
Advanced

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

Re: Octave 3.4 and Nan/Inf issue


From: Marco Atzeri
Subject: Re: Octave 3.4 and Nan/Inf issue
Date: Fri, 21 Oct 2011 18:38:00 +0200
User-agent: Mozilla/5.0 (Windows NT 5.1; rv:7.0.1) Gecko/20110929 Thunderbird/7.0.1

On 10/21/2011 6:16 PM, octaveuser001 wrote:
Hi,
I am using the new Octave 3.4.2 for windows (after following all
instructions including disabling certain packages as illustrated in the web
page).
I have a script - which if i run more than once, octave is behaving
strangely with errors. The actual function which is complaining is SVD
(saying cannot take SVD of matrix containing Inf or Nan values).
But i think the problem is elsewhere while i am computing dot products /
some multiplications and the matrix values are getting set to Nan/Inf.

  Note that if i close octave, open it again and run the script again, for
the first time it all works perfectly and i am getting my expected output.
Every other executions (sometimes from the 3rd time, sometimes from the 4th
time) - it gives error - i did try with clear all after a execution as well.
All these in command line - no GUI/debug complications.

  And here is another thing - i can run the same script 10 times in 3.2 - and
it works like a charm.

  I have no clue where some of the entries in the matrix are set to inf/Nan.
The code uses deterministic inputs and doesnt depend on external content.
One specific line
A1 = ones(m, 1) * (R * [-1 0 0]')';
seemed to behave strangely once when i put a disp for its contents - but i
couldnt repro it again.

  Has anyone noticed anything similar to this? I know this is an abstract
question but if someone has noted a similar behavior as above and have some
diagnostic tips - please let me know.



how we can try to replicate the problem without your script?

as SVD uses external libraries, it could be a problem there

Marco



reply via email to

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