The authors use an immune evolutionary programming to design multilayer feed-forward networks in this paper. The immune evolutionary programming retains the ability of stochastic global searching of traditional evolutionary programming, and draws into the interaction mechanism based on density and the diversity maintaining mechanism which exists in living beings' immune procedure. The immune evolutionary programming has better global convergence and very strong self-adaptive ability with enviornment. The experimental results prove the high efficiency of the immune evolutionary programming in designing neural networks.