Nature Inspired Computing Algorithms Performance and Applications

Abstract

Nature Inspired algorithms have served as the backbone of modern computing technology and over the past three decades, the eld has grown enormously. A large number of applications have been solved by these algorithms and are replacing the traditional classical optimization algorithms. In this thesis, some of these nature inspired algorithms such as cuckoo search algorithm (CS), ower pollination algorithm (FPA) and others have been studied. All these algorithms are state-of-the-art algorithms and have proven their worth in terms of competitiveness and application to various domains of research. The aim is to develop new improved algorithms through mitigating well-known problems that these algorithms su er from, such as local optima stagnation, poor exploration, slow convergence and parametric complexity. Such improvements should help these new variants to solve highly challenging optimization problems in contrast to existing algorithms. Di erent ideas and logic are employed in designing such new versions such as hybridization that combine the strength of di erent mutation strategies to add diversity in the solution space, adaptive parameter adaptations to converge faster, improved global and local search strategy to enhance the exploration and exploitation respectively. Also self-adaptivity, population size reduction and lower computational complexity methods have been analysed to provide prospective algorithms for the next generation researchers. Apart from these, based on the mating patterns of naked mole-rat, a new algorithm namely naked mole-rat algorithm (NMR) was proposed. To validate the performance of all these developed algorithms, various challenging test suites from the IEEE-CEC benchmarks are used. Each of these benchmarks constitute problems of di erent characteristics such as ruggedness, multimodality, noise in tness, ill-conditioning, non-separability and interdependence.