Design of enhanced deep belief network based on APSO

Abstract The deep belief network (DBN) has been widely applied in various fields. However, the traditional DBN has low prediction accuracy and it is difficult to determine the network structure when processing the continuous data. To solve these problems, an enhanced continuous deep belief network (EDBN) based on adaptive particle swarm optimization (APSO) is proposed in this paper. First, based on the gaussian noise transformation and deep learning theory, a EDBN is presented to improve the prediction accuracy of the deep belief network. Second, an APSO with adaptive mutation strategy is employed to optimize the network structure. Finally, the performance of EDBN-APSO is verified by comparing with other algorithms on Lorentz time series, concrete compressive strength test and the total phosphorus prediction. The results demonstrate the effectiveness of the proposed EDBN.