The connection between information theory and object search in networks

Abstract

This paper aims to establish a connection between information theory and object search in networks. We make two contributions:1) Using information theoretical arguments, we establish fundamental lower bounds on lookup table sizes and search step numbers. These bounds generalize those derived previously and can deal with non-uniform lookup distributions and object popularities. 2) Using the analogy between search sequences and data compression and coding, we propose a distributed implementation of Shannon code that can reduce the expected length of search sequences to an arbitrarily small value at the cost of at most doubling the table sizes.