machine-learningscikit-learnmultilabel-classificationscikit-multilearn

What is the difference between "OneVsRestClassifier" (Scikit-learn) and "Binary Relevance" (Scikit-multilearn)?

In scikit-learn, there is a strategy called sklearn.multiclass.OneVsRestClassifier, which can be used for both multiclass and multilabel problems. According to its documentation:

"In the multilabel learning literature, OvR is also known as the binary relevance method".

My question is,

Is there is any difference between this scikit-learn strategy and skmultilearn.problem_transform.BinaryRelevance?

Thank you in advance.

Solution

  • No, there is no difference. They work in the exact same way. They both predict probabilities for an instance belonging to a class, yes or no.