Training method of spiking neural network, image classification method, program, medium and equipment, and spiking neural network

The invention provides a training method of a spiking neural network, an image classification method, a program, a medium, equipment and the spiking neural network. And the pulse neural network is used for classifying the input image. In the feature extraction process, according to the training method of the spiking neural network provided by the embodiment of the invention, a feature mixing module including Gaussian difference, convolution and position embedding is adopted, so that the feature extraction capability of the model is enhanced; in addition, a scale factor is adopted in the attention mechanism, so that the image classification accuracy is improved. 本发明提供了一种脉冲神经网络的训练方法及图像分类方法、程序、介质和设备以及脉冲神经网络。所述脉冲神经网络用于对输入图像进行分类。在特征提取的过程中，根据本发明实施例的脉冲神经网络的训练方法通过采用了包括高斯差分、卷积和位置嵌入的特征混合模块，增强了模型的特征提取能力；此外，通过在注意力机制中采用了比例因子，提高了图像分类准确性。