Visualization to Assist Interpretation of the Multilevel Paradigm in Bipartite Graphs

Autor(es)

Cintra, Diego S. ; Valejo, Alan ; Lopes, Alneu A. ; Oliveira, Maria Cristina F. ; Kerren, A ; Hurter, C ; Braz, J

• Tipo de documento: Artigo Científico
• Fonte: VISAPP: PROCEEDINGS OF THE 15TH INTERNATIONAL JOINT CONFERENCE ON COMPUTER VISION, IMAGING AND COMPUTER GRAPHICS THEORY AND APPLICATIONS, VOL 4: VISAPP; v. N/A, p. 8-pg., 2020-01-01.

Resumo

Multilevel methods refer to a general framework for solving optimization problems in large graphs considering a hierarchy of contracted representations of the target graph. A recent extension to bipartite graphs has been introduced and successfully employed in diverse applications, but experience suggests the method is highly susceptible to the choice of vertex matching strategy for graph contraction and on whether the supervertices are relevant generalizations to the problem addressed. Although the flexibility in obtaining contracted representations of an original graph is a potential advantage, appropriate choice and parameterization of the contracting algorithms is challenging. Experts would benefit from solutions capable of assisting them in assessing alternatives and making informed decisions. In this work we describe a visualization system that creates an interactive graphical representation of a multilevel graph hierarchy obtained as a result of executing a multilevel method on bipartite graphs. We provide illustrative case studies showing the proposed visualization can support algorithm developers in inspecting and interpreting how different parameter choices in the coarsening stage impact the resulting multilevel hierarchies. (AU)

• Processo FAPESP: 16/25107-0 - Visualização de grandes grafos apoiada pelo paradigma multinível e aglomeração de arestas
• Beneficiário: Diego Silva Cintra
• Modalidade de apoio: Bolsas no Brasil - Mestrado
• Processo FAPESP: 17/05838-3 - Visual analytics: aplicações e uma investigação conceitual
• Beneficiário: Maria Cristina Ferreira de Oliveira
• Modalidade de apoio: Auxílio à Pesquisa - Regular