Abstract:

The cuckoo search algorithm utilizes Lévy Flights random walk and Biased random walk iteratively to search for the dimensional information of each new individual. After generating all dimensional information of each individual, the algorithm combines them as an individual and then evaluates it. Thus, the individuals with good dimensional information can be abandoned because of dimensional interference, which affects the convergence speed and refining capacity. We propose an enhanced cuckoo search algorithm  based on the strategy of dimension by dimension evaluation to accept those with good dimensional information, which can enhance the convergence speed and refining capacity. In the proposed approach, the strategy of dimension by dimension evaluation is adopted as a local search technique. We embed it behind the two random walks, and evaluate new individuals after each dimension is generated from those randomly selected individuals. Experimental results show that the dimension by dimension evaluation strategy can generally and effectively improve the performance of solution and the convergence speed.

Key words: cuckoo search algorithm, dimension by dimension evaluation, dimension by dimension improvement, local searc