help-octave
[Top][All Lists]
Advanced

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

Re: Wrong results for "Large" simultaneous equation calculation


From: Ben Abbott
Subject: Re: Wrong results for "Large" simultaneous equation calculation
Date: Sat, 16 Jan 2010 22:36:57 -0500

On Jan 16, 2010, at 1:09 PM, David Parman wrote:
>> 
>> On Fri, Jan 15, 2010 at 4:20 PM, Ben Abbott <address@hidden> wrote:
>> On Jan 15, 2010, at 5:58 PM, David Parman wrote:
>> 
>> > I am new to Octave, and ran into a problem with my first real application. 
>> > I am working with a large (120x120) set of simultaneous equations. When I 
>> > solve the equations using the A\b nomenclature, I get an answer - no 
>> > complaints from Octave. But the answer is clearly very wrong - all of the 
>> > values are about 6 or more orders of magnitude smaller than they should 
>> > be. I have the same set of calculations in MathCad, which yields an 
>> > appropriate solution. I can also approximate the solution using direct 
>> > analytical approach, which also confirms that the Octave result is way off.
>> >
>> > I found a couple of examples online for this type of solution (small 2x2 
>> > and 3x3 examples, one of which had the wrong answer on the webpage - 
>> > confirmed by Octave and MathCad). These all worked just fine, but they 
>> > have "nice" values (some of my values can be very large numbers, on the 
>> > order of 10^18).
>> >
>> > I have spot checked the matrices and the values appear to be exactly the 
>> > same in Octave as they are in MathCad. Does anybody have an idea what I 
>> > might be missing here?
>> 
>> Unfortunately if you don't give us an example that doesn't work, no one will 
>> be able to help.
>> 
>> Fortunately, you can easily provide your example :-)
>> 
>> If the matrix is only 120x120, try  ...
>> 
>>        save example.mat A b
>> 
>> ... and attach example.mat to your reply.
>> 
>> Ben
> 
> Ben,
> 
> Attached is the file. Any advice you can give will be greatly appreciated.


The error looks quite reasonable to me.

        octave:111> norm (A*(A\b)-b) / max (b)
        ans =  8.0399e-14

I don' t see how Octave's solution can be off by 6 orders of magnitude ... 
hmmm, I'm running the developers sources, what version of Octave are you 
running?

Ben

p.s. please <reply-all> and respond at the bottom so that others may follow 
along.





reply via email to

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