NEURAL NETWORK MODEL OPTIMIZATION PROCESSING METHOD AND DEVICE

摘要

The present disclosure provides a neural network model optimization processing method and device. The method comprises determining at least one target optimization layer of a neural network model on the basis of the number of neurons of each layer in the neural network model; for each target optimization layer in the at least one target optimization layer, adding a virtual layer between the target optimization layer and the upper layer of the target optimization layer, and the neurons of the virtual layer are respectively connected with the neurons of the target optimization layer and the upper layer of the target optimization layer, wherein the addition of the virtual layer enables the number of connections between the target optimization layer and the upper layer of the target optimization layer to be reduced; and training the neural network model with the added virtual layer to obtain an optimized neural network model. The device comprises a target optimization layer determination unit, a virtual layer addition unit and a model training unit. The neural network model is effectively compressed on the basis of a mathematically proven equivalent lossless compression method, the number of parameters of the model during training and use is reduced, and the storage space is saved.