Abstract: Multi-population leader dominating genetic algorithm (MPLDGA) is presented to overcome the shortcomings of the simple genetic algorithm (SGA). In the MPLDGA, the common individuals are crossed with the leader and the muti-population evolution strategy is adopoted, which make the MPLDGA with merits of fast convergence rate, robust global searching ability and less time-cost. Firstly, the algorithm is adopoted to optimize the excitation cooefficients of a 18-element linear array, and a cosecant-squared pattern ranging from -5° to 25° is realized with the side-lobe levels better than -30 dB. Finally, it is adopoted to optimize a 401-element planar sparse array, and the side-lobe levels are better than -21.9 dB at a thinning percentage of 60.8%.