deep-learning - Max pool layer vs Convolution with stride performance

Question

In most of the architectures, conv layers are being followed by a pooling layer (max / avg etc.). As those pooling layers are just selecting the output of previous layer (i.e. conv), can we just use convolution with stride 2 and expect the similar accuracy results with reduced process need?

Answer 1

是的，这是可以做到的。它在论文'Striving for simplicity: The all convolutional net' https://arxiv.org/pdf/1412.6806.pdf中进行了解释。引用论文：

“我们发现，在几个图像识别基准测试中，最大池化可以简单地用一个步幅增加的卷积层代替，而不会损失准确性”