简体繁体 English

如何解释这个 CNN 架构

[英]How to interpret this CNN architecture

原文 2022-04-27 02:49:38 1 1 python/ tensorflow/ conv-neural-network

How does this CNN architecture work from an input layer to the first convolution layer?这种 CNN 架构如何从输入层到第一个卷积层工作？ hx98 are input matrix dimensions, is n the number of channels or the number of inputs? hx98 是输入矩阵维度，n 是通道数还是输入数？

It doesn't seem like n is the number of channels because 25 is the number of feature maps and their dimensions do not indicate they are two channels.看起来 n 不是通道数，因为 25 是特征图的数量，并且它们的尺寸并不表示它们是两个通道。

However if n is the number of inputs and matrices are single channel, I haven't found a single CNN architecture anywhere that takes multiple input matrices and convolute them together.但是，如果 n 是输入的数量并且矩阵是单通道的，那么我在任何地方都没有找到一个单一的 CNN 架构，它需要多个输入矩阵并将它们卷积在一起。 Most example convolute them seperately and then concatenate.大多数示例将它们分别卷积然后连接。

In my example, n is 2, one is matrix with BER values and another with connection line-rate values.在我的示例中，n 为 2，一个是具有 BER 值的矩阵，另一个是具有连接线速率值的矩阵。

What mistake am I making?我犯了什么错误？ How does this CNN work.这个 CNN 是如何工作的。

1 个解决方案

In CNN the image pixels with height and width are multiplied with the kernel weights of the convolution layer and are added to create a feature map.在 CNN 中，具有高度和宽度的图像像素与卷积层的核权重相乘并相加以创建特征图。
The kernel will pass through all the channels of the image (3 channels for RGB, 1 channel for GreyScale) based on the strides defined in the convolution layer.内核将根据卷积层中定义的步幅通过图像的所有通道（RGB 为 3 个通道，GreyScale 为 1 个通道）。
After the convolution, the size of the image is reduced.卷积之后，图像的尺寸减小了。
To get the same output dimension as the input dimension, you need to add padding.要获得与输入维度相同的输出维度，您需要添加填充。 Padding consists of adding the right number of rows and columns on each side of the matrix.填充包括在矩阵的每一侧添加正确数量的行和列。 For details, please refer to this documentation .有关详细信息，请参阅此文档。 Thank You.谢谢你。