深度学习——Bottleneck Layer

Bottleneck Layer // Bottleneck  Features // Bottleneck Block

输入输出维度差距较大，就像一个瓶颈一样，上窄下宽亦或上宽下窄

"In order to reduce the number of weights, 1x1 filters are applied as a "bottleneck" to reduce the number of channels for each filter" 1x1 filters 可以起到一个改变输出维数（channels）的作用

在ResNet中：

其对应的图示如下：

右图中 1x1 filters把维度（channels）升高了，输入输出维度差距较大。继续如下图所示：

Bottleneck features are generated from a multi-layer perceptron in which one of the internal layers has a small number of hidden units, relative to the size of the other layers

 

转自：https://blog.youkuaiyun.com/u011501388/article/details/80389164