Currently betweenness is computed with the edge weights corresponding to trade.
IMHO, it's not quite right, because trade corresponds more to flow than cost , so for the computation of the shortest paths we should be using 1/weight, i.e., inverse of flow = cost.
This is already how we compute efficiency; betweenness was implemented somewhat absent-mindedly.
This PR fixes this.
Currently betweenness is computed with the edge weights corresponding to trade. IMHO, it's not quite right, because trade corresponds more to flow than cost , so for the computation of the shortest paths we should be using 1/weight, i.e., inverse of flow = cost. This is already how we compute efficiency; betweenness was implemented somewhat absent-mindedly. This PR fixes this.