Optimizing Extremely Randomized Trees with GPUs

In collaboration with Kenil Tanna.

Introduction

Random Forests are often the textbook example of highly parallel algorithms that are unsuitable for GPUs. In this project, we propose and evaluate our CUDA implementation of Extremely Randomized Trees (or ExtraTrees), a variant of Random Forests which are more amenable to GPU-driven parallel programming.

Report

You can find a copy of our report here.

zphang/extratrees_cuda

Optimizing Extremely Randomized Trees with GPUs

Introduction

Report