卷积层和下采样层一般会取若干个，采用卷积层和下采样层交替设置，即一个卷

Convolution layer and the lower layer will generally take the sample number, and the lower layer was sampled convolution layers are alternately arranged, i.e. a connection layer of a convolution downsampling layer, a connecting layer after downsampling convolution layer, and so on. Due to the convolution output of each neuron layer as its input connected to local surface characteristics, and by summing the weighted corresponding value of the local input connection weights plus a bias value, obtained input value of the neural element

Convolution layer and subsampling layer will generally take several, using the convolution layer and the subsampling layer alternate setting, that is, a convolution layer is connected to a subsampling layer, the subsampling layer is then connected to a convolution layer, and so on. The input value of the neuron is obtained because each neuron in the convolution layer has a local connection to its input and is weighted sum and bias edifyby by the corresponding connection weight and local input.

The convolution layer and the lower sampling layer are generally selected, and the convolution layer and the lower sampling layer are alternately set, that is, a convolution layer is connected with a lower sampling layer, and then the lower sampling layer is connected with a convolution layer, and so on. Each neuron in the output feature plane of convolution layer is locally connected with its input, and the input value of the neuron is obtained by the weighted sum of the corresponding connection weight and the local input plus the bias value<br>