Fig. 6From: Partitioning multi-layer edge network for neural network collaborative computingCommunication overhead for different paths. We set up an edge network with 10–100 edge devices to test the path selection ability of Ant Colony Algorithm by comparing with other three methods, i.e., Greedy Algorithm, Minimum Forwarding Algorithm, Random Selection. It shows that the cost of the paths increases with the number of the devices for all four methods. However, Ant Colony Algorithm can select a better path than other methods. Compared with the Greedy Algorithm and the Minimum Forwarding Algorithm, the Ant Colony Algorithm can jump out of the local optimal solution to better select the path with less communication overheadBack to article page