卷积神经网络 (convolutional neuron networ

Convolutional neuron networks (CNN) consists of one or more convolutional layers and a fully connected layer at the top, and includes relevant weights and a pooling layer. This structure enables CNN to use input data Two-dimensional structure. Compared with other deep structures, convolutional neural networks have shown excellent results in image and speech applications. Convolutional neural networks can also be trained using standard back-propagation algorithms, and because there are fewer parameter estimates, it is easier to train than other deep structures. The <br>figure is a schematic diagram of a convolutional neural network. The input image is convolved with three trainable filters. After convolution, three feature maps are generated in the Cl layer. After the feature map weights and offsets, the three S2 layers are obtained through a sigmoid function. Feature map. These maps are filtered to get the C3 layer. This hierarchical structure produces S4 like S2. Finally, these pixel values ​​are rasterized and connected into a vector input to the traditional neural network, and finally get the output. <br>Among them, the C layer is the feature extraction layer. The input of each neuron is connected to the local receptive field of the previous layer, and the local features are extracted. Once the local feature is extracted, its positional relationship with other features Determined; the S layer is a down-sampling layer, and each computing layer of the network is composed of multiple feature maps, each feature map is a plane, and the weights of all neurons on the plane are equal. The feature map structure uses the sigmoid function with a small influence function core as the activation function of the convolutional network, so that the feature map has displacement invariance

Convolutional networks (CONvolutional neuronal networks, CNN) consist of one or more convolutional layers and the top of the fully connected layers, and include related weights and pooling layers, a structure that enables CNN to leverage the two-dimensional structure of input data. Compared with other deep structures, convolutional neural networks show excellent results in image and speech applications. Convolutional neural networks can also be trained using standard reverse propagation algorithms and are easier to train than other depth structures due to the fact that there are fewer parameter estimates<br>The diagram of the convolutional neural network is shown as shown. The input image is convolued through three trained filters, which are then convoluible to produce three feature maps at the Cl layer, and then the feature map is weighted and biased, and the feature map of the three S2 layers is obtained through a sigmoid function. These maps then get the C3 layer through the filtered wave. This hierarchy then produces S4 in the same way as S2. Eventually, these pixel values are grated and connected into a vector input into a traditional neural network, which is finally output.<br>Wherein, the input of each neuron of the C layer is connected to the local sensory field of the previous layer, and the local feature is extracted, and once the local feature is extracted, the position relationship between it and other features is determined; The feature mapping structure uses the sigmoid function, which affects the smaller core of the function, as the activation function of the convolution network, making the feature mapping invariable in displacement

The convolutional neural network (CNN) is composed of one or more convolution layers and full connectivity layers at the top, and includes correlation weights and pooling layers. This structure enables CNN to use the two-dimensional structure of input data. Compared with other deep structures, convolutional neural network shows excellent results in image and speech applications. Convolutional neural network can also be trained by using standard back propagation algorithm, and it is easier to train than other depth structures due to its less parameter estimation<br>As shown in the figure is the schematic diagram of convolutional neural network. The input image is convoluted with three trainable filters. After convolution, three feature maps are generated in CL layer. Then the weighted value and offset of the map are used to get three S2 layer feature maps through a sigmoid function. These maps are filtered to obtain C3 layer. This hierarchy again generates S4 like S2. Finally, these pixel values are rasterized and connected into a vector input to the traditional neural network, and finally the output is obtained.<br>Among them, layer C is the feature extraction layer, the input of each neuron is connected with the local receptive field of the previous layer, and the local feature is extracted. Once the local feature is extracted, the location relationship between it and other features is determined; layer s is the downsampling layer, and each calculation layer of the network is composed of multiple feature maps, each feature map is a plane, The weights of all neurons in the plane are equal. The sigmoid function with smaller influence function kernel is used as the activation function of convolution network in the structure of feature mapping, which makes the feature mapping have displacement invariance<br>