The low-rank eigenvalue problem

Abstract: The nonzero eigenvalues of

A B

are equal to those of

B A

: an identity that holds as long as the products are square, even when

A, B

are rectangular. This fact naturally suggests an efficient algorithm for computing eigenvalues and eigenvectors of a low-rank matrix

X = A B

with

A, B^{T} i n m a t h b b C^{N i m e s r}, N g g r

: form the small

r i m e s r

matrix

B A

and find its eigenvalues and eigenvectors. For nonzero eigenvalues, the eigenvectors are related by

A B v = l a m b d a v L e f t r i g h t a r r o w B A w = l a m b d a w

with

w = B v

, and the same holds for Jordan vectors. For zero eigenvalues, the Jordan blocks can change sizes between

A B

and

B A

, and we characterize this behavior.

This page was built for publication: The low-rank eigenvalue problem