Skip to content

Commit

Permalink
update conv-layer.md
Browse files Browse the repository at this point in the history
  • Loading branch information
panhaoyu committed Aug 24, 2023
1 parent e6b18cc commit b77d188
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion chapter_convolutional-neural-networks/conv-layer.md
Original file line number Diff line number Diff line change
Expand Up @@ -347,7 +347,7 @@ d2l.reshape(conv2d.weight, (1, 2))
请注意,感受野可能大于输入的实际大小。让我们用 :numref:`fig_correlation`为例来解释感受野:
给定$2 \times 2$卷积核,阴影输出元素值$19$的感受野是输入阴影部分的四个元素。
假设之前输出为$\mathbf{Y}$,其大小为$2 \times 2$,现在我们在其后附加一个卷积层,该卷积层以$\mathbf{Y}$为输入,输出单个元素$z$。
在这种情况下,$\mathbf{Y}$上的$z$的感受野包括$\mathbf{Y}$的所有四个元素,而输入的感受野包括最初所有九个输入元素
在这种情况下,$z$在层$\mathbf{Y}$上的感受野包括$\mathbf{Y}$的所有四个元素,而在输入的$\math3 \times 3$的张量上的感受野包括最初所有九个输入元素
因此,当一个特征图中的任意元素需要检测更广区域的输入特征时,我们可以构建一个更深的网络。

## 小结
Expand Down

0 comments on commit b77d188

Please sign in to comment.