Low-cost search in scale-free networks

We study local search algorithms for networks with heterogeneous edge weights, testing them on scale-free and Erdos-Renyi networks. We assume that the location of the destination node is discovered when it is two edges away, and that the search cost is additive. It was previously shown that a search strategy preferring high-degree nodes reduces the average search cost over a simple random walk. In the prior work, for the case when the edge costs are randomly distributed, a different preference was investigated [high local betweenness centrality (LBC)], and was found to be superior to high-degree preference in scale-free networks, with the exception for the most sparse ones. We have found several preference criteria that are simpler and which, in all networks we tested, yield a lower cost than other criteria including high-degree, high-LBC, and low-edge cost.