Google PageRank How to increase your PR

Google PageRank How to increase your PR? Laure Ninove, Cristobald de Kerchove and Paul Van Dooren CESAME, Université catholique de Louvain, Belgium Golub memorial, Feb 29, 2008 Google sorts webpages according to their PRs Improving your PR will make your webpage more visible The PR equations PR vector π is obtained from a combination of a random walk on the graph, with probability c and a preferential zapping, with probability (1-c) The PR equations PR vector π is obtained from a combination of a random walk on the graph, with probability c and a preferential zapping, with probability (1-c) Stochastic matrix: Dii = outdegree(i) A = adjacency matrix The PR equations PR vector π is obtained from a combination of a random walk on the graph, with probability c and a preferential zapping, with probability (1-c) Damping factor c in ]0,1[ Stochastic matrix: Dii = outdegree(i) A = adjacency matrix The PR equations PR vector π is obtained from a combination of a random walk on the graph, with probability c and a preferential zapping, with probability (1-c) Damping factor c in ]0,1[ Stochastic matrix: Dii = outdegree(i) A = adjacency matrix Personalization vector z > 0 and zT e =1. The Google matrix G is the left eigenvector of The Google matrix G is the left eigenvector of G is irreducible and stochastic, therefore is the stationary distribution of the corresponding Markov chain The Google matrix G is the left eigenvector of G is irreducible and stochastic, therefore is the stationary distribution of the corresponding Markov chain G is the transition probability matrix How to improve your PR? How to improve your PR? adding any in-link and removing well-chosen outlinks How to improve your PR? Add inlinks How to improve your PR? Add inlinks 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 How to improve your PR? Add inlinks you 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 How to improve your PR? Add inlinks you 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 How to improve your PR? Add inlinks you 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 You have 3 inlinks. Adding new inlinks, improves always your PR. How to improve your PR? Add inlinks How to improve your PR? Choose outlinks How to improve your PR? Choose outlinks you 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 You have 2 outlinks. Adding/Removing new outlinks, does not always improve your PR. How to improve your PR? Choose outlinks Optimal choice of outlinks for a single node for a set of nodes Optimal choice of outlinks for a single node for a set of nodes Optimal choice of outlinks for a single node The Google matrix : you = a single node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 Optimal choice of outlinks for a single node The Google matrix : you = a single node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 1 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 FIXED 0 1 0 0 0 0 1 0 0 0 1 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 We maximize the PR of your single webpage by choosing the outlinks. Optimal choice of outlinks for a single node Proposition : the maximal PR for node 1 is obtained by having only one outlink to a particular parent of node 1. But some parents may be less interesting than nodes that are not a parent of node 1. Optimal choice of outlinks for a single node Optimal choice of outlinks for a single node You must choose the nodes with smallest mean return time to your node; this can be expressed in terms of G Optimal choice of outlinks for a single node for a set of nodes Optimal choice of outlinks for a set of nodes The Google matrix : you = a single node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 Optimal choice of outlinks for a set of nodes The Google matrix : you = a set of node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 Optimal choice of outlinks for a set of nodes The Google matrix : you = a set of node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 You have 6 outlinks. You have 5 inlinks. Optimal choice of outlinks for a set of nodes The Google matrix : you = a set of node 0 1 1 1 0 0 0 0 1 0 0 1 0 1 0 1 0 0 1 1 0 0 1 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 FIXED 0 0 1 0 0 0 1 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 1 1 0 1 0 0 1 0 1 0 0 1 0 0 0 We maximize the sum of PRs of the set of nodes by choosing the outlinks. Optimal choice of outlinks for a set of nodes Proposition : the maximal sum of PRs is obtained by having k outlinks under the constraint that the set of nodes must have at least k outlinks (in order not to be dismissed by Google) Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. Optimal choice depends on G and does not depend explicitly on z ! Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. 25 FIXED 25 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED zoom 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED 99 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. FIXED Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. Optimal local choice Monotonic function Random FIXED 99 Optimal choice of outlinks for a set of nodes Idea : it is always possible to remove an outlink in such a way that the sum of PRs increases. The optimum is not always reached by the monotonic or local optimal choice Optimal local choice Monotonic function Random 99 Concluding remarks We modify the outlinks of Concluding remarks We modify the outlinks of Results do not extend to maximization of cTπ FIXED Concluding remarks We modify the outlinks of Results do not extend to maximization of cTπ FIXED What about modifying its internal structure ? Optimal structure is knwon but complex Adding a link between 2 pages of can decrease the PR of one of these pages or even, can decrease the sum of their PRs ! FIXED What about Gene’s graph ? What about Gene’s graph ? The hub-authority algorithm used to rate « authorities » is nothing but the SVD …

Related docs
pagerank
Views: 131  |  Downloads: 7
The Value of Pagerank
Views: 43  |  Downloads: 0
Digg url with high pagerank for backlink
Views: 226  |  Downloads: 8
The_5_Myths_About_Google_Pagerank
Views: 5  |  Downloads: 0
Get to the Top on Google.pdf
Views: 107  |  Downloads: 18
Google_Inc.
Views: 3  |  Downloads: 2
Google AdSense A to Z
Views: 26  |  Downloads: 6
premium docs
Other docs by thebest11
Manufacturers business plan
Views: 506  |  Downloads: 20
Notice of Directors Meeting
Views: 145  |  Downloads: 1
Juniper Networks Inc Ammendments and Bylaws
Views: 188  |  Downloads: 1
brown-all
Views: 355  |  Downloads: 1
CORPORATION DISSOLUTION INFORMATION CHECKLIST
Views: 413  |  Downloads: 4
LoisLawcom Inc Ammendments and Bylaws
Views: 201  |  Downloads: 1
Fluor Corp Ammendments and Bylaws
Views: 177  |  Downloads: 0
r494
Views: 280  |  Downloads: 2
CMGI Inc Ammendments and By laws
Views: 257  |  Downloads: 1
Employee Monthly Time Record
Views: 336  |  Downloads: 10
Customer Bounced Check Letter
Views: 4244  |  Downloads: 47