zmllovemln / ssa-dbn-classification Goto Github PK

Combining the advantages of deep belief network (DBN) in extracting features and processing high-dimensional and non-linear data, a classification method based on deep belief network is proposed. This method uses the Fourier spectrum (FFT) of the original time domain signal to train a deep confidence network through deep learning. Its advantage is that the method does not need to set parameters when performing FFT on the signal, and directly uses all spectral components for modeling, so there is no need for complexity The feature selection method has strong versatility and adaptability. Finally, in order to further enhance the classification accuracy of DBN, the Sparrow Search Algorithm (SSA) is used to optimize the weight parameters of DBN. The experimental results show that the method proposed in this paper can effectively improve the classification and recognition accuracy.