Cuckoo search algorithm is a novel stochastic global optimization algorithm based on swarm intelligence,with advantages of few control parameters,optimal search path and good global search capability,but it also has shortcomings of weak local search ability,slow convergence velocity and low convergence accuracy.In order to overcome these disadvantages of CS algorithm,an improved cuckoo search algorithm based on conjugate gradient is introduced.After evolved from Levy flights and elimination mechanism,the cuckoo populations decline rapidly in the mutually conjugate directions so that the convergence ability of algorithm is strengthened significantly under the condition of maintaining the strong global search capability of CS algorithm.The CGCS algorithm and CS algorithm are tested by four typical test functions.The conclusions indicate that CGCS algorithm has faster convergence velocity,higher convergence accuracy and more stable optimization results.Meanwhile CGCS algorithm has good global search capability,convergence ability and robustness,which is particularly suitable for the optimization of multimodal function and high dimension function.