The phylogenetic tree of boosting has a bushy carriage but a single trunk
Richard Nock, Frank Nielsen · Proceedings of the National Academy of Sciences · 2020
The phylogenetic tree of boosting has a bushy carriage, with early influencers: Ref. 1 was undoubtedly one for the making and use of popular application packages, some used nowadays in almost every Kaggle competition (2), and in that respect it is an understatement to say that the recent work of Luna et al. (3), embedded in the framework of interpretability, is important. There is, however, a single trunk to this tree, and it grows from roots 1) in the probably approximately correct (PAC) model of Valiant (4), which led to 2) opening the boosting problem (5)—namely, how to use an oracle returning “weak” classifiers slightly better … [↵][1]1To whom correspondence may be addressed. Email: Richard.Nock{at}data61.csiro.au. [1]: #xref-corresp-1-1