Few-shot learning 기본 개념(Basic concepts)

기본적인 컨셉

Support set

• Support set is a small set of smaples → It is too small for training a model.
• Every class has at most a few samples.
• The support set can only provied additional information at test time.

Query Set

• Query samples are never seen before.
• Query samples are from unkonwn classes.

Traditional supervised learning

• Test samples are never seen before.
• Test samples are from known classes.

 

Few-shot learning

• Few-shot learning is the problem of making predictons based on a limited number of samples.
• The goal of few-shot learning is not to let the model recognize the images in the training set and then generalize to the test set.

Meta learning

• The goal is learn to learn.
• The goal is to know the similarity and difference between objects.
• The model has learned the similarity and difference between objects.

 

기본 개념 정리

• Few-shot learning is a kind of meta learning.
• Meta learning: learn to learn
• Traditional supervised learning asks the model to recognize the training data and then genralize to unseeen test data.

 

K-way n-shot Support Set

• K-way: the support set has k classes
• n-shot: every class has n samples

 

 

way의 수가 늘어날수록, 예측 정확도는 떨어집니다.

 

 

3가지 보단 6가지일 때의 경우가 더 어렵다고 짐작이 갑니다.

 

shots의 수가 늘어날수록, 예측 정확도는 올라갑니다.

 

 

samples 수가 늘어날수록 정확도가 올라간다고 생각할 수 있습니다.

 

 

자료 출처: https://youtu.be/hE7eGew4eeg