Figure 3

The channel model highlights the directed flow from origins to destinations. Consider once again the example network from Figures1 and2, now with a single source (hexagon) and two sinks (octagons). In common with the case from Figure2, the walkers start at the source, but in this case can terminate only by reaching the sinks. The damping factor is implicit: it determines how far the walkers are allowed to deviate from the shortest path towards one of the sinks. In the first five graphs, the number in each transient node documents the total number of visits to that node from all random walkers up to the indicated time. However, the value in each sink node represents the likelihood to reach that sink from the source at the indicated time. The value of${\widehat{\Phi }}_{i,K}^s$ returned by the ITM Probe normalized channel mode yields the expected number of visits to node i per random walker that starts at s over infinitely many time steps. Note that the sink nodes split the flow from the source depending on their location. In this example, over infinitely many time steps, the node closer to the source captures 970 walkers, while the farther sink gets only the remaining 30.