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

Sampling Matters in Deep Embedding Learning #314

Open
chullhwan-song opened this issue Feb 26, 2020 · 1 comment
Open

Sampling Matters in Deep Embedding Learning #314

chullhwan-song opened this issue Feb 26, 2020 · 1 comment

Comments

@chullhwan-song
Copy link
Owner

@chullhwan-song
Copy link
Owner Author

chullhwan-song commented Mar 10, 2020

Abstract

  • 이 연구에서는 loss 관련 논문

  • 기존 loss 연구들은 손실(loss) 역할을 하는 부분에 맞춰있지만, 훈련 셋에서 샘플링하는 방법도 그와동등하게 매우 중요하다.는 것을 보여준다.

  • distance weighted sampling & margin based loss
    image

  • distance weighted sampling

    • 좁은 지역에서 샘플링대신에 퍼진(spread) 넓은 영역에서 샘플링 : inverse distance > image
      • anchor와 negative sampling 사이에서..
      • This gives us examples which are spread out instead of being clustered around a small region.
    • 이러한 특성에 대한 다른 loss와의 비교
      image
  • Margin based loss
    image

    • 베타는 pos 와 neg사이의 boundary 결정하는 요인
    • 알파는 분리되는 margin을 control
      image
  • 실험
    image
    image

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

No branches or pull requests

1 participant