Skip to yearly menu bar Skip to main content


Poster

Universal consistency and minimax rates for online Mondrian Forests

Jaouad Mourtada · Stéphane Gaïffas · Erwan Scornet

Pacific Ballroom #12

Keywords: [ Boosting and Ensemble Methods ] [ Regression ] [ Classification ] [ Online Learning ] [ Learning Theory ]


Abstract: We establish the consistency of an algorithm of Mondrian Forests~\cite{lakshminarayanan2014mondrianforests,lakshminarayanan2016mondrianuncertainty}, a randomized classification algorithm that can be implemented online. First, we amend the original Mondrian Forest algorithm proposed in~\cite{lakshminarayanan2014mondrianforests}, that considers a \emph{fixed} lifetime parameter. Indeed, the fact that this parameter is fixed actually hinders statistical consistency of the original procedure. Our modified Mondrian Forest algorithm grows trees with increasing lifetime parameters $\lambda_n$, and uses an alternative updating rule, allowing to work also in an online fashion. Second, we provide a theoretical analysis establishing simple conditions for consistency. Our theoretical analysis also exhibits a surprising fact: our algorithm achieves the minimax rate (optimal rate) for the estimation of a Lipschitz regression function, which is a strong extension of previous results~\cite{arlot2014purf_bias} to an \emph{arbitrary dimension}.

Live content is unavailable. Log in and register to view live content