Abstract: Cuckoo search is a novel meta-heuristic algorithm based on bionics. It has good ability to search for global optimum，but suffers from slow searching speed in the last iterations and poor accuracy.The cross-entropy method is embedded into cuckoo search algorithm and an improved cuckoo search algorithm is introduced. The testing results of benchmark functions show the improved algorithm obtains good performance of convergence speed and high accuracy. The proposed algorithm is employed as a new training method for artificial neural network.，and the experimental results show that the proposed algorithm outperforms for training neural networks in terms of converging speed and avoiding local minima. Finally，the artificial neural network with learning algorithm based on the improved algorithm is employed to forecast total population of China.

Key words: artificial neural network, cuckoo search（ CS ）, cross entropy（ CE ）, Chinese population forecasting