In mathematics, power iteration (also known as the power method) is an eigenvalue algorithm: given a diagonalizable matrix , the algorithm will produce a number , which is the greatest (in absolute value) eigenvalue of , and a nonzero vector , which is a corresponding eigenvector of , that is, . The algorithm is also known as the Von Mises iteration.[1]

Power iteration is a very simple algorithm, but it may converge slowly. The most time-consuming operation of the algorithm is the multiplication of matrix by a vector, so it is effective for a very large sparse matrix with appropriate implementation. The speed of convergence is like (see a later section). In words, convergence is exponential with base being the spectral gap.

The method

edit
 
Animation that visualizes the power iteration algorithm on a 2x2 matrix. The matrix is depicted by its two eigenvectors. Error is computed as  

The power iteration algorithm starts with a vector  , which may be an approximation to the dominant eigenvector or a random vector. The method is described by the recurrence relation

 

So, at every iteration, the vector   is multiplied by the matrix   and normalized.

If we assume   has an eigenvalue that is strictly greater in magnitude than its other eigenvalues and the starting vector   has a nonzero component in the direction of an eigenvector associated with the dominant eigenvalue, then a subsequence   converges to an eigenvector associated with the dominant eigenvalue.

Without the two assumptions above, the sequence   does not necessarily converge. In this sequence,

 ,

where   is an eigenvector associated with the dominant eigenvalue, and  . The presence of the term   implies that   does not converge unless  . Under the two assumptions listed above, the sequence   defined by

 

converges to the dominant eigenvalue (with Rayleigh quotient).[clarification needed]

One may compute this with the following algorithm (shown in Python with NumPy):

#!/usr/bin/env python3

import numpy as np

def power_iteration(A, num_iterations: int):
    # Ideally choose a random vector
    # To decrease the chance that our vector
    # Is orthogonal to the eigenvector
    b_k = np.random.rand(A.shape[1])

    for _ in range(num_iterations):
        # calculate the matrix-by-vector product Ab
        b_k1 = np.dot(A, b_k)

        # calculate the norm
        b_k1_norm = np.linalg.norm(b_k1)

        # re normalize the vector
        b_k = b_k1 / b_k1_norm

    return b_k

power_iteration(np.array([[0.5, 0.5], [0.2, 0.8]]), 10)

The vector   converges to an associated eigenvector. Ideally, one should use the Rayleigh quotient in order to get the associated eigenvalue.

This algorithm is used to calculate the Google PageRank.

The method can also be used to calculate the spectral radius (the eigenvalue with the largest magnitude, for a square matrix) by computing the Rayleigh quotient

 

Analysis

edit

Let   be decomposed into its Jordan canonical form:  , where the first column of   is an eigenvector of   corresponding to the dominant eigenvalue  . Since generically, the dominant eigenvalue of   is unique, the first Jordan block of   is the   matrix   where   is the largest eigenvalue of A in magnitude. The starting vector   can be written as a linear combination of the columns of V:

 

By assumption,   has a nonzero component in the direction of the dominant eigenvalue, so  .

The computationally useful recurrence relation for   can be rewritten as:

 

where the expression:   is more amenable to the following analysis.

 

The expression above simplifies as  

 

The limit follows from the fact that the eigenvalue of   is less than 1 in magnitude, so

 

It follows that:

 

Using this fact,   can be written in a form that emphasizes its relationship with   when k is large:

 

where   and   as  

The sequence   is bounded, so it contains a convergent subsequence. Note that the eigenvector corresponding to the dominant eigenvalue is only unique up to a scalar, so although the sequence   may not converge,   is nearly an eigenvector of A for large k.

Alternatively, if A is diagonalizable, then the following proof yields the same result

Let λ1, λ2, ..., λm be the m eigenvalues (counted with multiplicity) of A and let v1, v2, ..., vm be the corresponding eigenvectors. Suppose that   is the dominant eigenvalue, so that   for  .

The initial vector   can be written:

 

If   is chosen randomly (with uniform probability), then c1 ≠ 0 with probability 1. Now,

 

On the other hand:

 

Therefore,   converges to (a multiple of) the eigenvector  . The convergence is geometric, with ratio

 

where   denotes the second dominant eigenvalue. Thus, the method converges slowly if there is an eigenvalue close in magnitude to the dominant eigenvalue.

Applications

edit

Although the power iteration method approximates only one eigenvalue of a matrix, it remains useful for certain computational problems. For instance, Google uses it to calculate the PageRank of documents in their search engine,[2] and Twitter uses it to show users recommendations of whom to follow.[3] The power iteration method is especially suitable for sparse matrices, such as the web matrix, or as the matrix-free method that does not require storing the coefficient matrix   explicitly, but can instead access a function evaluating matrix-vector products  . For non-symmetric matrices that are well-conditioned the power iteration method can outperform more complex Arnoldi iteration. For symmetric matrices, the power iteration method is rarely used, since its convergence speed can be easily increased without sacrificing the small cost per iteration; see, e.g., Lanczos iteration and LOBPCG.

Some of the more advanced eigenvalue algorithms can be understood as variations of the power iteration. For instance, the inverse iteration method applies power iteration to the matrix  . Other algorithms look at the whole subspace generated by the vectors  . This subspace is known as the Krylov subspace. It can be computed by Arnoldi iteration or Lanczos iteration. Gram iteration[4] is a super-linear and deterministic method to compute the largest eigenpair.

See also

edit

References

edit
  1. ^ Richard von Mises and H. Pollaczek-Geiringer, Praktische Verfahren der Gleichungsauflösung, ZAMM - Zeitschrift für Angewandte Mathematik und Mechanik 9, 152-164 (1929).
  2. ^ Ipsen, Ilse, and Rebecca M. Wills (5–8 May 2005). "7th IMACS International Symposium on Iterative Methods in Scientific Computing" (PDF). Fields Institute, Toronto, Canada.{{cite news}}: CS1 maint: multiple names: authors list (link)
  3. ^ Pankaj Gupta, Ashish Goel, Jimmy Lin, Aneesh Sharma, Dong Wang, and Reza Bosagh Zadeh WTF: The who-to-follow system at Twitter, Proceedings of the 22nd international conference on World Wide Web
  4. ^ Delattre, B.; Barthélemy, Q.; Araujo, A.; Allauzen, A. (2023), "Efficient Bound of Lipschitz Constant for Convolutional Layers by Gram Iteration", Proceedings of the 40th International Conference on Machine Learning