Lec 29 | 18.06 Linear Algebra, Spring 2005

[ysn1216]

I've confronted the same problem below. I've just followed Prof. and I got in trouble. I'm not sure but the sigma matrix has to be sqrt(32) and -sqrt(18). Acutally (sigma1)^2=32 and (sigma2)^2=18 so that we cannot say sigma1=sqrt(32) and sigma2=sqrt(18) because they could be minus.

[death0intj]

i had this problem with SVD, since v and -v is the same eigenvector my jacobi routine returned valid data but in SVD it does matter which one i'll use, so i ended up getting wrong decomposition (maybe that's the cause of the error in first example on that lecture) i ended up calculating only 'U' and finding 'V' from (U,S and M)

[mazemaster225]

SVD is so good. Sphere to ellipsoid, big pimpin. So many proofs become trivial if you just consider the SVD.

[itsthebrod]

Terrible camerawork (did they keep falling asleep?!) but great lecture!

[KurtlarOlmez]

especially for nonsquare matrices, svd provides a better way to calculate pseudo inverse of the given matrix.

[KurtlarOlmez]

svd is a very important method

i have used this technique in calculating inertia matrices for a humanoid robot

thank you professor