π
, where the ith
entry is the PageRank score for webpage i
, is an eigenvector for the eigenvalue λ=1
of G
.
LetFor small Google matrices, we can quickly find exact solutions to the eigensystem,A=(ajk)
be ann×n
matrix andX
a column vector. For the equationAX=λX
, the roots of this equation are called eigenvalues of the matrixA
. Corresponding to each eigenvalue, there will be a solutionX≠0
, which is called an eigenvector for the eigenvalue.
πG=π
.
The Google matrix for the entire Web has more than 25 billion rows and columns, so computing the exact solution requires extensive time and computing resources.
The oldest and easiest technique for approximating an eigenvector of a matrix is the power method:
Given a starting vectorFor example, the PageRank vector of the Model 1 is found as follows:π(0)
, e.g.,π(0)=v
, the power method calculates successive iteratesπ(k) = π(k-1)until some convergence criterion is satisfied.G
, wherek=1,2,...,
3/80 71/80 3/80 3/80 π(0)G = vG = [ 1/4 1/4 1/4 1/4 ] × [ 3/80 3/80 71/80 3/80 ] 37/80 3/80 3/80 37/80 1/4 1/4 1/4 1/4 = [ 0.197 0.303 0.303 0.197 ] = π(1) 3/80 71/80 3/80 3/80 π(1)G = [ 0.197 0.303 0.303 0.197 ] × [ 3/80 3/80 71/80 3/80 ] 37/80 3/80 3/80 37/80 1/4 1/4 1/4 1/4 = [ 0.208 0.247 0.341 0.208 ] = π(2) ... π(k) = [ 0.21 0.26 0.31 0.21 ] ≈ π (PageRank vector)