An artificial ants model for fast construction and approximation of proximity graphs

Abstract

In this paper we present a summary of our work which has led to the conception of a new model for the fast construction of proximity graphs. We present the state of the art in graph self-assembly, and then we detail the self-assembly behavior observed in real ants from which our model is derived. We describe our main algorithm, called AntGraph, where each ant represents one datum and where the proximity graph is built in an incremental way. Ants perform two steps: following the path of maximum local similarity, and then connecting to other ants. We present a hierarchical extension, called H-AntGraph, which can build large graphs (with up to 1 million data items). We study the properties of the constructed graphs, and compare our results with those obtained by other methods. We use force-directed graph layout algorithms to display the graphs and to allow the domain expert to perform an interactive clustering task. We validate this approach with a comparative user-study.