Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Education and Culture Executive Agency (EACEA). Neither the European Union nor EACEA can be held responsible for them.

Library

Xgboost: A scalable tree boosting system

Users: 3 - Average Rating: 4.33


In this paper, tree boosting algorithm is described, which is a very popular and effective solution widely recognized in several machine learning and data mining competitions. The main reason for the success of XGBoost is its scalability in different scenarios: it runs 10 times faster than the existing popular solutions and scales beyond billions of examples using far fewer resources than existing systems. In particular, the authors introduce a novel sparsity-aware algorithm for handling sparse data and weighted quantile sketch for approximate tree learning which both contribute to the scalability of XGBoost; more importantly, they provide insights on cache access patterns, and out-of-core computations (data compression and sharding) to build a scalable tree boosting system.
Different datasets have been considered to evaluate the scaling property of the system and the impact of out-of-core computations and sparsity-aware algorithm.

Type:
Scientific Paper

Area:
Data Analytics, Machine Learning

Target Group:
Advanced

DOI:
https://doi.org/10.1145/2939672.2939785

Cite as:
Chen, T. and Guestrin, C., Xgboost: A scalable tree boosting system, Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (2016): 785-794.

Author of the review:
Giulia Cademartori
University of Genoa


Reviews

You have to login to leave a comment. If you are not registered click here

Andrej Košir


A very flexible classifier applicable to non-linear spaces with relatively simple configuration parameters. If the classification space is complex and poorly understood, is a good way to start with.

Mauro Bozzetti


Great Resource!!

Joana Lopes


Very Useful