Understanding the Power Iterative Method to find eigenvalues

300 Views Asked by At

I'm slightly confused about how to use the power method and the steps to calculate an eigenvalue. - I understand that the power method is defined as U(x+1) = AU(x)/a(x) where "a" is the first component of U(x). I do not understand at all what "U" is. Are we picking any vector we want that minimizes the error? What would I do given the practice problem below?

Apply the power method to $$ \begin{matrix} 1 & 2 \\ 3 & -4 \\ \end{matrix} $$

to obtain three approximations of the largest eigenvalue of A. What is the limiting vector u∞?

1

There are 1 best solutions below

3
On BEST ANSWER

The key is, that $U$ - whatever initial value you take - converges to the first eigenvector (the eigenvector to the largest eigenvalue). So you simply start with $U=[1,0,0,...0]$ and iterate.

There is one caveat: the largest eigenvalue must (in absolute value) be greater than 1 and distinct from the second highest.

Here is a table of iterations. The actual vector U is horizontal written as entry 1 and 2, and the current approximation to the first eigenvalue is in the third column. The first(largest in absolute value) eigenvalue is negative, so the system needs a certain time to reduce oscillations:

  U[1]            U[2]             eigenvalue
 -----------------------------------------
  1                  0                 0
  1                  3                 1
  1               -9/7                 7
  1             -57/11             -11/7
  1           -261/103           -103/11
  1          -1353/419          -419/103
  1         -6669/2287         -2287/419

in float reprecentation

  U[1]            U[2]     eigenvalue
 -----------------------------------------
  1.0000000           0           0
  1.0000000   3.0000000   1.0000000
  1.0000000  -1.2857143   7.0000000
  1.0000000  -5.1818182  -1.5714286
  1.0000000  -2.5339806  -9.3636364
  1.0000000  -3.2291169  -4.0679612
  1.0000000  -2.9160472  -5.4582339
  1.0000000  -3.0347480  -4.8320944
  1.0000000  -2.9862913  -5.0694960
  1.0000000  -3.0055137  -4.9725827
  1.0000000  -2.9977994  -5.0110274
  1.0000000  -3.0008810  -4.9955987