A Communication-Efficient Parallel Algorithm for Decision Tree

TL;DR

This paper introduces PV-Tree, a parallel decision tree algorithm that minimizes communication costs by using local and global voting, enabling scalable and efficient training with high accuracy.

Contribution

The paper presents PV-Tree, a novel parallel decision tree algorithm that reduces communication overhead and maintains near-optimal learning performance.

Findings

PV-Tree significantly reduces communication costs compared to existing methods.

PV-Tree achieves comparable or better accuracy on real-world datasets.

Theoretical analysis confirms PV-Tree's ability to find the best attribute with high probability.

Abstract

Decision tree (and its extensions such as Gradient Boosting Decision Trees and Random Forest) is a widely used machine learning algorithm, due to its practical effectiveness and model interpretability. With the emergence of big data, there is an increasing need to parallelize the training process of decision tree. However, most existing attempts along this line suffer from high communication costs. In this paper, we propose a new algorithm, called \emph{Parallel Voting Decision Tree (PV-Tree)}, to tackle this challenge. After partitioning the training data onto a number of (e.g., $M$) machines, this algorithm performs both local voting and global voting in each iteration. For local voting, the top-$k$ attributes are selected from each machine according to its local data. Then, globally top-$2k$ attributes are determined by a majority voting among these local candidates. Finally, the full-grained histograms of the globally top-$2k$ attributes are collected from local machines in order to identify the best (most informative) attribute and its split point. PV-Tree can achieve a very low communication cost (independent of the total number of attributes) and thus can scale out very well. Furthermore, theoretical analysis shows that this algorithm can learn a near optimal decision tree, since it can find the best attribute with a large probability. Our experiments on real-world datasets show that PV-Tree significantly outperforms the existing parallel decision tree algorithms in the trade-off between accuracy and efficiency.