An efficient dynamic resource allocation algorithm for packet-switched communication networks based on Hopfield neural excitation method

Dynamic Resource Allocation (DRA) algorithms set up different connections over the same resources and perform a scheduling policy to distribute the resources usage. Recently, intelligent DRA techniques based on Hopfield Neural Networks computational methods have been proposed, showing their potential for solving this kind of complex optimization problems. However, the initial algorithms suffer from severe instability problems impacting performance. This paper addresses these specific limitations stressing the proper neuron dynamics and proposing an efficient energy formulation and an optimum calculation of the weighting coefficients. These changes result in a maximum resource utilization together with an optimized neural network convergence.