deeplearning.ai 总结 -1x1 convolutional layer(1x1卷积层有什么作用)
flyfish
在Inception Network Motivation 有讲解
没有使用1×1的情况
计算量
(
28
×
28
×
32
)
×
(
5
×
5
×
192
)
≈
120
m
i
l
l
i
o
n
(28×28×32)×(5×5×192)≈ 120 million
(28×28×32)×(5×5×192)≈120million
使用1×1的情况
计算量
(
28
×
28
×
16
)
×
(
1
×
1
×
192
)
+
(
28
×
28
×
32
)
×
(
5
×
16
×
16
)
≈
12.4
m
i
l
l
i
o
n
(28×28×16)×(1×1×192)+(28×28×32)×(5×16×16)≈ 12.4 million
(28×28×16)×(1×1×192)+(28×28×32)×(5×16×16)≈12.4million
1x1 convolutional layer的作用
Let’s say you have a 28 by 28 by 192 volume. If you want to shrink the height and width, you can use a polling layer. So we know how to do that. But one of a number of channels has gotten too big and we want to shrink that. How do you shrink it to a 28 by 28 by 32 dimensional volume? Well, what you can do is use 32 filters that are one by one. And technically, each filter would be of dimension 1 by 1 by 192, because the number of channels in your filter has to match the number of channels in your input volume, but you use 32 filters and the output of this process will be a 28 by 28 by 32 volume.
如果你有一个 28 ∗ 28 ∗ 192 28*28*192 28∗28∗192 输入层,若要压缩它的高度和宽度,可以通过池化层(pooling layer),如果channel数目太大,怎么将将它压缩到 28 ∗ 28 ∗ 32 28*28*32 28∗28∗32 你可以使用 32 32 32个滤波器 1 ∗ 1 1*1 1∗1(filter/卷积核),每个滤波器的大小都是 1 ∗ 1 ∗ 192 1*1*192 1∗1∗192,因为滤波器中channel的数量必须与输入层channel的数量保持一致。