❗

version space 를 정의할 때 노이즈(혹은 outlier) 는 전혀 고려되지 않는 대상인걸까

상태

✅ 해결된 궁금증

생성일

2022/03/25 09:52

관련태그

1 more property

전혀 고려되지 않는 대상이다 (참고1).

참고

1.

A major drawback of version space learning is its inability to deal with noise: any pair of inconsistent examples can cause the version space to collapse, i.e., become empty, so that classification becomes impossible.[1] One solution of this problem is proposed by Dubois and Quafafou that proposed the Rough Version Space,[3] where rough sets based approximations are used to learn certain and possible hypothesis in the presence of inconsistent data.