Neural Random Forests

TL;DR

This paper introduces neural random forests, a hybrid model combining decision trees and neural networks, offering improved flexibility, fewer parameters, and strong empirical performance on diverse datasets.

Contribution

It reformulates random forests as neural networks and proposes two new hybrid methods with theoretical guarantees and practical advantages.

Findings

Proven consistency of the proposed methods.

Strong empirical results on synthetic and real datasets.

Fewer parameters and less restrictive decision boundaries than traditional models.

Abstract

Given an ensemble of randomized regression trees, it is possible to restructure them as a collection of multilayered neural networks with particular connection weights. Following this principle, we reformulate the random forest method of Breiman (2001) into a neural network setting, and in turn propose two new hybrid procedures that we call neural random forests. Both predictors exploit prior knowledge of regression trees for their architecture, have less parameters to tune than standard networks, and less restrictions on the geometry of the decision boundaries than trees. Consistency results are proved, and substantial numerical evidence is provided on both synthetic and real data sets to assess the excellent performance of our methods in a large variety of prediction problems.