+
+ // The edges will be order as follow: First the nb_locations *
+ // nb_time_steps inside the node pairs, which will have lengths
+ // equal to the opposite of the detection scores, then the ones
+ // between node pairs, and finally the ones from source and to
+ // sink.