Skip to content

(장*우)(강의노트6장) GBRT 학습률과 트리 개수 관계 #178

@glok1d

Description

@glok1d

GBRT의 과대적합을 막기 위해 각 트리의 기여도를 줄이는 축소 규제(learning_rate)를 사용한다고 배웠습니다. 그렇다면 learning_rate를 0.1에서 0.01로 10배 줄였다면 원래 세팅해둔 트리의 개수(n_estimators)도 상식적으로 10배 늘려주는 것이 맞는지 두 파라미터가 서로 반비례하며 움직이는 구체적인 매커니즘이 궁금합니다.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions