Learning from Noisy Labels with Deep Neural Networks: A Survey

My Motivation

What is noise label?

: 잘못 어노테이션된 레이블들

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/e8445eb6-dc31-41d2-9304-d6eec07d8ce1/Screen_Shot_2021-03-30_at_9.47.42_AM.png

기존의 데이터에대한 생각

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/2363037c-2221-4bed-a54f-303af0574a54/Untitled.png

Robust Training에서 달성하고 싶은 목표

: 일부 노이즈한 레이블이 들어와도 노이즈에 강한 학습을 수행하자

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/43552cd7-5741-4ca5-a5dc-b3e88cca518f/Screen_Shot_2021-03-30_at_9.49.22_AM.png

Introduction

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/0a0af851-44de-47f3-8361-00675626484f/Screen_Shot_2021-03-30_at_9.50.30_AM.png

Preliminaries

Deep Learning Approaches