The report consists of three independent parts (one by each of the authors) which describe three different methods of finding clusters (i.e.,strongly interconnected sets of nodes) in graphs having weighted arcs. The first method deletes nodes that are connected weakly to the remainder of the graph,and takes the connected components of what survives to be clusters. The second method deletes arcs at each node that are significantly weaker than the local average arc weight;it too takes the connected components of what survives to be the clusters. The third method recomputes the weight assigned to each arc after examining all paths of length 2between its nodes,then thresholds the modified weights;with this method, a variety of cluster definitions can be used. (Author)
展开▼
