Comparison-Based Random Forests
Séminaire Probabilités & Statistique
7/02/2019 - 14:00 Mr Damien Garreau (Max Planck Institute for Intelligent systems Statistical Learning Theory group) Salle 106 - Batiment IMAG
Le lien pour le code est disponible ici : https://github.com/SiavashCS/CompRF Le lien vers l'article correspondant à la présentation est le suivant : http://proceedings.mlr.press/v80/haghiri18a Abstract: Assume we are given a set of items from a general metric space, but we neither have access to the representation of the data nor to the distances between data points. Instead, suppose that we can actively choose a triplet of items (A, B, C) and ask an oracle whether item A is closer to item B or to item C. In this paper, we propose a novel random forest algorithm for regression and classification that relies only on such triplet comparisons. In the theory part of this paper, we establish sufficient conditions for the consistency of such a forest. In a set of comprehensive experiments, we then demonstrate that the proposed random forest is efficient both for classification and regression. In particular, it is even competitive with other methods that have direct access to the metric representation of the data. Joint work with Siavash Haghiri and Ulrike von Luxburg