CO2 Forest: Improved Random Forest by Continuous Optimization of Oblique Splits
This addresses the need for more accurate classification models in machine learning, though it is incremental as it builds on existing Random Forest and oblique tree techniques.
The paper tackles the problem of improving Random Forest classifiers by optimizing multivariate linear threshold functions for splits, resulting in forests of up to 1000 trees that significantly outperform standard Random Forest and previous oblique tree methods on benchmarks like LFW.
We propose a novel algorithm for optimizing multivariate linear threshold functions as split functions of decision trees to create improved Random Forest classifiers. Standard tree induction methods resort to sampling and exhaustive search to find good univariate split functions. In contrast, our method computes a linear combination of the features at each node, and optimizes the parameters of the linear combination (oblique) split functions by adopting a variant of latent variable SVM formulation. We develop a convex-concave upper bound on the classification loss for a one-level decision tree, and optimize the bound by stochastic gradient descent at each internal node of the tree. Forests of up to 1000 Continuously Optimized Oblique (CO2) decision trees are created, which significantly outperform Random Forest with univariate splits and previous techniques for constructing oblique trees. Experimental results are reported on multi-class classification benchmarks and on Labeled Faces in the Wild (LFW) dataset.