Power Method: Theory

CS357 Fall 2016

Andrew Liang
aliang6
November 11, 2016

1
Observe that for a shift matrix and generic vector of size n,
x x
0 1 0 0 1
x x
0 ...
1 2


S~x = x2 = ..

(1)
. . .

. . 1

.. xn1
1 0 xn1 x0

So applying S once shifts or rotates the coordinates of a vector up by one index. Then we can reasonably
assume that if we shift a vector that has already been shifted once, the resulting vector will be the original
vector shifted up by two indices. We generalize this to the case of Sk ~x for 0 k < n, where the resulting
vector will have its components shifted or rotated upwards by k indices like so:

xk
xk+1
..

.
Sk ~x =

xn1 (2)

x0

.
..
xk1

2
Unfortunately, the power method does not work on the shift matrix. Note that (2) can be re-written for
values of k N assuming ~x is normalized

xk mod n
x(k+1) mod n
Sk ~x = (3)

..
.
x(k+n1) mod n

Which does not oscillate or stabilize as k due to the modular nature of the matrix.