Abstract This paper presents a fast global optimization neural network and applies it to the data fusion. This neural network is based on the global property of genetic algorithm and the high speed property of expectation maximization (EM) algorithm. The simulation results show that this neural network is robust in the data fusion.