Propagation Connectivity of Random Hypergraphs
We consider the average-case MLS-3LIN problem, the problem of finding a most likely solution for a given system of perturbed 3LIN-equations generated with equation probability p and perturbation probability q. Our purpose is to investigate the situation for certain message passing algorithms to work for this problem whp We consider the case q = 0 (no perturbation occurs) and analyze the execution of (a simple version of) our algorithm. For characterizing problem instances for which the execution succeeds, we consider their natural 3-uniform hypergraph representation and introduce the notion of “propagation connectivity”, a generalized connectivity property on 3-uniform hypergraphs. The execution succeeds on a given system of 3LIN-equations if and only if the representing hypergraph is propagation connected. We show upper and lower bounds for equation probability p such that propagation connectivity holds whp on random hypergraphs representing MLS-3LIN instances.
KeywordsRandom Graph Initial Assignment Equation Probability Propagation Connectivity Marked Vertex
Unable to display preview. Download preview PDF.
- [Ale03]Alekhnovich, M.: More on average case vs approximation complexity. In: Proc. FOCS 2003, pp. 298–307. IEEE, Los Alamitos (2003)Google Scholar
- [CCF09]Coja-Oghlan, A., Cooper, C., Frieze, A.M.: An efficient sparse regularity concept. In: Proc. SODA 2009, pp. 207–216 (2009)Google Scholar
- [OW09]Onsjö, M., Watanabe, O.: Finding most likely solutions. In: Theory of Computing Systems. On-Line Version (2009)Google Scholar
- [WY07]Yamamoto, M., Watanabe, O.: Belief propagation and spectral methods. Research Report C-248, Dept. of Math. Comput. Sci., Tokyo Inst. of Tech. (2007)Google Scholar