首先将特征图矩阵补一圈零, 然后用权重矩阵对补零后的特征图矩阵进行 v

First, the feature map matrix is ​​filled with a circle of zeros, and then the weighted matrix is ​​used to perform a vaild-type convolution operation on the zero-filled feature map matrix. Here, the number of zero-filling layers is equal to the size of the convolution kernel matrix minus 1, full-type convolution The connotation is to make the convolution start from the edge,

First, the feature chart matrix is filled with a circle of zeros, and then the feature chart matrix after the weight matrix is used to carry out the vaild convolution operation. The number of layers that fill the zero here is equal to the size of the convolution core matrix minus 1, and the full convolution is meant to make the convolution start from the edge.

First, the eigengraph matrix is filled with a circle of zeros, and then the eigengraph matrix after zero filling is convoluted with the weight matrix in the form of shield. Here, the number of layers of zero filling is equal to the size of the convolution kernel matrix minus 1. The connotation of full convolution is to make the convolution start from the edge,<br>