Bom dia a todos, Teremos no dia 20 de Outubro, sexta-feira, das 10:30 às 12:00, a discussão da tese de mestrado do Miguel Duarte, "Fast hierarchical graph clustering". Resumo: This work studies how graph clustering methods can be used to solve other graph problems, such as hierarchical clustering and graph reordering. In this context, it explores how Faststep, a recent method for matrix factorization and graph clustering, can be adapted to solve these other problems. Our study develops a new method for this task, based on Faststep, and compares the proposed algorithm with state-of-the-art clustering methods: Layered Label Propagation and Louvain method. The results obtained are evaluated by several tests, which include comparison with ground-truth communities, networks generated by random models and direct comparison of algorithms. This study takes also advantage of Webgraph, a framework for succinct representation of graphs, and explores how clustering can be a powerful tool to improve the compression rates. We conclude that Faststep, although designed to solve a different problem, can still obtain acceptable results for this one. We also show that Louvain method, a hierarchical clustering method, can in fact obtain really promising results in the graph reordering task. Saudações, Alexandre Francisco