Delay Constrained Multicast Tree with Neural Networks in Fixed Time

This study aimed to propose an artificial neural networks, in fixed time, for the construction of the delay constrained multicast tree. These neural networks are composed mainly of two kinds of neurons: linear and threshold neurons, which are easier to implement in a specialized hardware. Multicast routing consists of transmitting information from a single source to multiple destinations, using the network resources very effectively, and respecting several constraints, such as delay, cost, bandwidth or other. To guarantee an optimal diffusion, it is necessary to determine a tree that connects the source node to all destination nodes minimizing the use of resources. Our approach for solving this problem differs from the conventional approach used in the field of neural networks. Our primary concern is how to organize neurons into a network so that it can solve a specific problem, with an emphasis on fully utilizing the massive parallelism property offered by neural networks.