Networking device and method using a neural network for applying delays based on port positions. The networking device comprises a plurality of ports and applies a delay to IP packets received via a given port among the plurality of ports. The delay is based on a position of the given port among the plurality of ports. The neural network has been trained to infer the delay to be applied to the IP packet received via the given port. In one implementation, the position of the given port is one of the inputs of the neural network.