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

[week2] Projection Layer가 정확히 뭘까요? #10

Closed
cow-coding opened this issue Mar 17, 2022 · 2 comments
Closed

[week2] Projection Layer가 정확히 뭘까요? #10

cow-coding opened this issue Mar 17, 2022 · 2 comments

Comments

@cow-coding
Copy link
Collaborator

  • 본 논문에서 training complexity를 계산하는 수식에서 Q가 등장합니다. Q는 모델마다 다르게 정의하기로 하는데, 이 수식 연산에서 자주 등장하는 것이 projection layer (matrix) 입니다.
  • 정확하게 projection layer(matrix)가 뭔지 모르겠네요....?
    가장 비슷한건 embedding layer (matrix)가 떠오르는데 정확히 embedding이란 용어를 쓰진 않았으니...
@TB2715
Copy link
Collaborator

TB2715 commented Mar 17, 2022

Projection layer가 NNLM 에서 나온 개념인 것 같습니다.
보통 NN에서 input layer와 output layer 사이의 층은 hidden layer라고 하는데, projection layer는 일반 hidden layer와 다르게 가중치 행렬과의 곱셈은 이루어지지만 활성화 함수가 존재하지 않는다고 되어 있네요!

출처
출처2

@kimcando
Copy link
Owner

저는 projection 의미를 선형대수랑 연결해서 이해합니다.
matrix를 곱해주는 linear operation을 취하면 인풋을 아웃풋 공간으로 보내주는데요. 이때 인풋 공간에서 아웃풋으로 '보내주는' 것을 projection이라고 볼 수 있습니다. 즉 매트릭스를 곱해줌으로써 새로운 공간으로 보내주는 것( nxm이면 n차원 인풋을 m 차원으로 보내주는 것)을 projection이라하고, 실제로 메트릭스를 곱이 이루어지는 layer를 projection layer라고 보면 다른 논문들에서 projection 표현하는걸 이해하기 쉬울 것 같아요!

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

3 participants