Das Institut für Informatik lädt ein:

ZUM THEMA

An Algebraic Multigrid Solution of
Large Hierarchical Markovian Models

Prof. Dr. Udo R. Krieger

Faculty Information Systems and Applied Computer Science

Otto-Friedrich-University Bamberg

Wann: Mittwoch, 11.06.2014 um 13 Uhr s.t.

Wo: Hörsaal H 34, (AI)

ABSTRACT

In Web information retrieval stochastic link analysis provides important supplementary means to generate a ranking of searched objects. Considering a hierarchical algebraic description of a Web graph with host-oriented clustering of pages or a role-oriented perspective, we propose an efficient computation of the stationary distribution of the underlying homogeneous Markov chain of a random surfer by iterative aggregaton/disaggregation procedures and algebraic multigrid methods.

In particular, we descuss the application of an efficient multigrid variant of the multiplicative Schwarz iteration which can be performed on a single machine with limited storage requirements.

PDF öffnen/downloaden

Universität Bayreuth -