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

<In progress> Density estimation using Real NVP #11

Open
Yagami360 opened this issue Sep 8, 2019 · 0 comments
Open

<In progress> Density estimation using Real NVP #11

Yagami360 opened this issue Sep 8, 2019 · 0 comments

Comments

@Yagami360
Copy link
Owner

Yagami360 commented Sep 8, 2019

0. 論文情報・リンク

1. どんなもの?

  • フローベースの生成モデルにおいて、カップリング層として加法カップリング層 [additive coupling layer] ではなくアフィンカップリング層 [affine couping layer] 採用し、また Multi-scale architecture を採用するなどして、同種の先行研究である NICE よりも大幅に品質の高い画像生成を実現したフローベースの生成モデル。

2. 先行研究と比べてどこがすごいの?

  • 同種の先行研究である NICE より、生成画像の品質が大幅に優れている。

3. 技術や手法の"キモ"はどこにある?

  • affine coupling layer を採用:
    先行研究の NICE では、カップリング層として additive coupling layer を採用していたが、本手法では、以下の図と式で定義される affine coupling layer を採用している。
    image
    image
    これにより、ヤコビアンの計算は以下の式のように簡略化される。
    image

  • Multi-scale architecture の採用 :
    image
    xxx

  • Masked convolution :
    xxx

  • カップリング層の結合 :
    xxx

  • Batch normalization の使用:
    xxx

4. どうやって有効だと検証した?

  • xxx

5. 議論はあるか?

  • xxx

6. 次に読むべき論文はあるか?

  • Glow: Generative Flow with Invertible 1×1 Convolutions
@Yagami360 Yagami360 changed the title Density estimation using Real NVP <In progress> Density estimation using Real NVP Sep 29, 2019
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