Answers are grounded in this paper and cite their sources.
We consider the problem of multiclass classification. Our main thesis is that a simple one-vs-all scheme is as accurate as any other approach, assuming that the underlying binary classifiers are ...
No takes yet. Share an insight, caveat, or question.
RifkinRyan et al. (2004) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: