Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Questions about Differentiable Neural Architecture Search #9

Open
buttercutter opened this issue May 31, 2021 · 2 comments
Open

Questions about Differentiable Neural Architecture Search #9

buttercutter opened this issue May 31, 2021 · 2 comments

Comments

@buttercutter
Copy link

image

问题 1:
https://www.bilibili.com/video/BV1C64y127Fv 时间: 10:26

θ 不是张量 (看时间:7:00),所以卷积层block 的输出 不等于 θ

那么 卷积层block 的输出 怎么生成 θ 呢 ?

Question 1:
https://youtu.be/D9m9-CXw_HY?t=626

θ is not a tensor (check video at t=7:00), so the convolutional block output is not equal to θ

So, how does the convolutional block generate θ ?

image

问题 2:
D-X-Y/AutoDL-Projects#99 (comment)GDAS 文献 的式子 (7) 里,如何在两个节点之间的多个平行 的连接线 进行反向传播的运作 ?

Question 2:
For D-X-Y/AutoDL-Projects#99 (comment) and equation (7) of GDAS paper , how to do backpropagation across multiple parallel edges between two nodes ?

@wangshusen
Copy link
Owner

  1. Theta 不是神经网络的输出,而是预先定义的变量,比如初始化为 9x1 的全 1 向量,然后用反向传播做更新。
  2. 你可以复习一下多元函数的链式法则。比如 a = f(x), b = g(x), c = h(a, b), 该如何求 c 关于 x 的导数。

@buttercutter
Copy link
Author

  1. 明白了,谢谢。
  2. 在正常的反向传播中,两个节点之间只有一条连接线。 但是在 GDAS 里,两个节点之间有多个平行的连接线 ?

image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants