Multiprocessor scheduling by mean field theory

The authors develop an optimization scheme based on mean field theory (MFT) to solve the task scheduling problem. The algorithm combines characteristics of the simulated annealing (SA) algorithm and the Hopfield neural network. The temperature behavior of MFT for the task scheduling problem is shown to possess a critical temperature below which an optimal solution may be achieved. The algorithm has been applied to various task graphs, and promising results have been obtained.<
>