Impact of decentralization on the performance of a Hopfield neural network-based DCA scheme

The impact of decentralizing a Hopfield neural network-based dynamic channel allocation (HNN-DCA) scheme is evaluated. In this scheme, the channel assignment is performed autonomously by every cell, based on information gathered from a set of neighboring cells rather than relying upon a central resource manager and availability of complete system information. Centralized schemes are technically undesirable or computationally prohibitive, or both, since large optimization areas impose increased signaling load and channel allocation time. Results show that a distributed HNN-DCA exhibits the advantages of distributed schemes-scalability and distributed signaling load and reduced allocation time, without significant loss in performance.