Distributional Generalization

2000년 도입 · 논문 4편에서 사용

Distributional Generalization is a type of generalization that roughly states that outputs of a classifier at train and test time are close as distributions, as opposed to close in just their average error. This behavior is not captured by classical generalization, which would only consider the average error and not the distribution of errors over the input domain.

출처: Distributional Generalization: A New Kind of Generalization

소개 논문: Distributional Generalization: A New Kind of Generalization