An improved cuckoo search algorithm with self-adaptive knowledge learning

Cuckoo search (CS) is a one of the most efficient evolutionary for global optimization and widely applied to solve diverse problems in the real world. Despite its efficiency and wide use, CS suffers from premature convergence and poor balance between exploitation and exploration. To cope with these issues, a new cuckoo search algorithm extension based on self-adaptive knowledge learning (I-PKL-CS) is proposed. In this study, learning model with individual history knowledge and population knowledge is introduced into the CS algorithm. Individuals constantly adjust their position by using historical knowledge and communicate with each other by using their own knowledge in the optimization process. In order to reduce complexity of the I-PKL-CS algorithm, the optimal learning model is selected to exploit the potential of individual knowledge learning and population knowledge learning by adopting threshold statistics learning strategy, which provides a good trade-off between the exploration and exploitation. The accuracy and performance of the proposed approach are evaluated by eighteen classic benchmark functions and CEC 2013 test suite. Statistical comparisons of the experimental results showed that the proposed I-PKL-CS algorithm made an appropriate trade-off between exploration and exploitation. Comparing the proposed I-PKL-CS with various CS algorithms, variants of differential evolution, and improved particle swarm optimization algorithms, the results demonstrate that I-PKL-CS is a competitive new type of algorithm.