A search space of graph motifs for graph compression: From Powergraphs to triplet concepts
Defended on December 17, 2019.
Power Graph Analysis is a lossless graph compression method aiming at reducing the visual complexity of a graph. The process is to detect motifs, cliques and bicliques, which enables the hierarchical clustering of nodes, the grouping of edges, and ultimately a graph reduced to these groups. This thesis exposes first the formalization of the Power Graph Analysis search space, using Formal Concept Analysis as a theoretical ground to express the compression process. Because the independent treatment of two motifs presents some caveats, we propose a unification framework, triplet concepts, which encode a more general motif for compression. Both Power Graph Analysis and the new approach have been implemented in Answer Set Programming (ASP), a logical formalism, and we present some applications in bioinformatics of these two approaches. This thesis ends on the presentation of an high-level specification and visualization environment for graph theory.