Abstract: To resolve the problem of no guidance about how to set values of numerical meta-parameters parameters and difficulty to achieve optimization of Restricted Boltzmann Machines, genetic algorithms are used to develop an automatic optimizing method named GA-RBM (Genetic Algorithm-Restricted Boltzmann Machine) for Restricted Boltzmann Machines' aided design. Based on the features of Restricted Boltzmann Machines and evaluation function, a genetic algorithm is designed and realizes a global search of satisfied structure and network's weights to determine the number of visible units and hidden units, and also initial weights. Firstly we introduced RBM parameters affect the ability of the model to extract features, studied the genetic algorithm and a combination of RBM training, then proposed RBM model of individual coding and fitness function is designed for RBM training can effectively achieve through genetic algorithm simultaneously search for RBM network structure and weights. The experiments were conducted on MNIST digits handwritten datasets. The results proved that this optimization reduces the dimension of visible units; improves the performance of feature extracted by Restricted Boltzmann Machines. The network optimized has good generalization performance and meets the demand of Restricted Boltzmann Machines' aided design.