HELP ME WIN A BET WITH MY FRIEND (Matlab and LDA)

Aerionne · **Joined:** Fri, 10 Dec 2010 05:54:45 UTC **Posts:** 2

I have a friend that is putting radar data into MatLab and then uses LDA to obviously reduce the demensionality of the image, however, in the next to last step when matlab produces the "weights" (which might also be considered eigenvalues, I'm not sure) it is giving him complex numbers that he will not be able to put back in matlab to obtain the reduced image. I think he knows how to fix them, we just want to know WHY matlab is giving us complex values.

I told him that I was going to solve this before him, please help me win!!!!!!!!!!!

Thanks!!! :wink:

Shadow · **Posted:** Fri, 10 Dec 2010 06:09:54 UTC

Aerionne wrote:

I have a friend that is putting radar data into MatLab and then uses LDA to obviously reduce the demensionality of the image, however, in the next to last step when matlab produces the "weights" (which might also be considered eigenvalues, I'm not sure) it is giving him complex numbers that he will not be able to put back in matlab to obtain the reduced image. I think he knows how to fix them, we just want to know WHY matlab is giving us complex values.

I told him that I was going to solve this before him, please help me win!!!!!!!!!!!

Thanks!!! :wink:

Ummmm....real matrices can have complex eigenvalues. Just consider $\begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix}$

Aerionne · **Joined:** Fri, 10 Dec 2010 05:54:45 UTC **Posts:** 2

Can you explain how? Isn't the only way to be multiplying that matrices by a complex number.

Shadow · **Posted:** Fri, 10 Dec 2010 06:22:40 UTC

Aerionne wrote:

Can you explain how? Isn't the only way to be multiplying that matrices by a complex number.

What? The eigenvalues are the roots of the characteristic polynomial. This one has characteristic polynomial x^2+1

...no real roots.

S.O.S. Mathematics CyberBoard

HELP ME WIN A BET WITH MY FRIEND (Matlab and LDA)

Who is online