What's the difference between using the inverse trick VS cholesky trick for solving generalized eigenvalue problem?

26 Views Asked by At

My math professor said once that solving the generalized eigenvalue problem

$$A\lambda = \lambda B v$$

There are two methods to use:

Method #1

Use cholesky decomposition

$$B = LL^T$$

Solve $Y$

$$AY = L$$

Take the inverse of $L^T$ and multiply it with $Y$

$$X = Y*(L^T)^{-1}$$

Then you solve the eigendecomposition

$$X\lambda = \lambda v$$

And now you found the eigenvalues $\lambda$ and the eigenvectors $v$

Method #2

Take the inverse of $B$ and solve the eigendecomposition

$$B^{-1}A\lambda = \lambda v$$

And now you found the eigenvalues $\lambda$ and the eigenvectors $v$

Question:

According to my math professor, method #1 is much more prefered than method #2.

My question is: why?