mldivide test doesnt work.

Hello,

For a research project I have to find the coefficients corrsponding to the value of importance of the previous days.

for example:

_a = rand(1024,4);

b= rand(1024,1);

t=a\b;

btest=a*t;_

where t would give me the coefficients. However when I do the test by computing btest and compare it with b, this gives me a completely other result. How is this possible? Am I doing something wrong or is this because b is calculated with approximations by matlab?

Thanks!

Answer 1

[Mischa Kim]

Derck, you are working with an overdetermined system: 1024 equations in 4 unknowns. Chances that the system results in an exact solution are slim.

Try, instead (to get confidence in MATLAB)

 a     = rand(1024,1024);
 b     = rand(1024,1);
 t     = a\b;
 btest = a*t;

 norm(b-btest,inf)
 ans =
     8.495634751248815e-13

Answer 2

[Image Analyst]

btest won't equal b. I don't know if that's what you're thinking, but it won't happen unless you have perfect data to begin with. You're doing an ordinary least squares solution so your solution with be a nice smooth analytical line that goes nicely between the random points. Your test arrays are particularly badly chosen - in fact I don' think you could have possibly chosen worse sets of data . You're going to have no sensible equations that map a bunch of random values into another bunch of random values with a simple weighted sum of the input values.