An approximate quantile algorithm that uses weights, such as gradient statistics, to identify candidate split points.
A weighted quantile sketch is an approximate split-finding technique for gradient-boosted trees. It uses gradient statistics as weights, favoring candidate quantiles that may reduce large prediction errors. section 5.4.3
Use the technique within XGBoost’s split-finding choices:
Select tree_method="approx" to generate weighted candidate split points from gradient information. section 5.4.3
Distinguish approx from tree_method="hist": approx creates new bins each iteration, while hist reuses bins across iterations. section 5.4.3
Treat histogram-based splitting as the faster alternative to scanning continuous values, with hist the default since XGBoost 2.0.0. section 5.4.3
Histogram speed comparison
A latitude split produces the same result with and without 256 bins, but the histogram version takes about 0.56 seconds instead of 46.9 seconds.
How Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI covers it
How deep it goes
Explains weighted candidates, histogram acceleration, and bin reuse, without deriving the sketch’s internal quantile construction.
Examples
Compares exact and 256-bin histogram splitting on latitude using matching thresholds and sharply different runtimes.
What you can do after
Evaluate XGBoost split-finding choices by comparing exact, approx, and hist methods and their bin-reuse behavior.
How it connects
Connects gradient boosting with decision trees, histograms, quantiles, gradient statistics, information gain, and XGBoost.
Ensemble Methods for Machine LearningA weighted quantile sketch is an approximate data-structuring algorithm for estimating quantile-based split points while accounting for example weights.
XGBoostused by · XGBoost is an open-source software library for efficient gradient-boosted decision-tree algorithms used in supervised machine learning.
This explainer was generated by Manning's knowledge graph from the text of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI. The section links point to the parts of the book it draws on.
