Performance evaluation of approximated artificial neural network (AANN) algorithm for reliability improvement

Approximated artificial neural network (AANN) is a meta-heuristics optimization algorithm mixing the features of approximating the computed combinatorial spectrum and ability of neural network to approximate the input from problem domain to the desired output. This paper proposes the use of an approximated artificial neural network (AANN) approach for the case of reliability when complex network design is considered. A mesh network of 256 nodes and hyper-tree network of 496 nodes are considered for evaluating the performance of AANN algorithm for improving reliability and minimizing cost for complex network. Since, evaluating reliability for complex network using formal approach requires substantial computational effort and time equivalent to NP-Hard. The work presented in this paper compares the performance of AANN algorithm with that of Monte Carlo simulation (MCS) and particle swarm optimization (PSO) for improving reliability and minimizing cost for complex network problems. The simulation results show that the performance of AANN algorithm is comparable to those of the mentioned algorithms and can be used to improve reliability and reduce the cost for complex network design problems when amount of complexity is relatively higher.