Skip to content

[34] VQ-VAE: Neural Discrete Representation Learning #34

@Dongwoo-Im

Description

@Dongwoo-Im

Links

한 줄 요약

  • VAE의 continuous latent variable 대신 discrete latent variable을 사용하는 VQ-VAE 제안
  • Discretisation 과정에 vector quantisation (VQ)을 적용함으로써, VAE의 posterior collapse 현상을 방지할 수 있다고 주장
    • VQ-VAE가 latent discretisation을 처음으로 제안한 것은 아니지만 VAE와 비슷한 성능을 보였다는 점에서 유의미함
    • Posterior collapse는 approximate posterior가 gaussian prior z와 같아지는 현상으로 확인됨 (정확한 표현은 아닐 수 있음)

선택 이유

  • 유명한 논문이어서 + DALL-E를 읽기 위해 선정

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions