Gradient Descent Optimization for Routing in Multistage Interconnection Networks

摘要

The design and development of interconnection networks is a critical issue in the field of communication switches, routers and multiprocessors. Here we consider the problem of routing in multistage interconnection networks (MIN). In our previously reported research we proposed a method to find the best path in a graph. In this paper, we use our algorithm which is based on neural network for routing in MINs. In this algorithm, we define a suitable energy function; the minimum of this function correspond to a valid route. By using gradient descent method, the energy is minimized at the convergence of neural network. Simulation results show that this method finds a valid path between source and destination and because neurons act in parallel, the performance is comparable with other approaches.