Efficient Implementation of Decision Forests

This chapter describes a variety of techniques for writing efficient, scalable, and general-purpose decision forest software. It will cover:-

- algorithmic considerations, such as how to train in depth first or breadth first order;

- optimizations, such as cheaply evaluating multiple thresholds for a given feature;

- designing for multi-core, GPU, and distributed computing environments; and

- various `tricks of the trade', including tuning parameters and dealing with unbalanced training sets.

Publisher  Springer
Springer-Verlag London 2013


> Publications > Efficient Implementation of Decision Forests