Pulse neural network compression method suitable for neuromorphic hardware

The invention belongs to the technical field of computer vision and machine learning, and particularly relates to a pulse neural network compression method suitable for neuromorphic hardware. According to the method, the statistical characteristics of the spiking neurons in the spiking neural network and the statistical data in the batch normalization layer in the pulse convolution channel are comprehensively utilized, the efficient model compression effect is achieved on the basis that excessive redundant structures are not additionally added, and the performance of the pruned model is improved by introducing sparsity in the model fine tuning stage. Compared with the traditional neural network pruning technology, the spiking neural network compression method suitable for the neuromorphic hardware has the advantages that the model volume of the spiking neural network can be effectively reduced; and the model calculation amount is reduced, and the performance loss of the spiking neural network after pruning is