Improve Ranking E ciency by Optimizing Tree Ensembles

Claudio Lucchese

Franco Maria Nardini

Salvatore Orlando

Ra aele Perego

Fabrizio Silvestri

Salvatore Trani

0 2 0 ISTI-CNR , Pisa 1 University Ca' Foscari of Venice 2 University of Pisa 3 Yahoo London

Learning to Rank (LtR) is the machine learning method of choice for producing highly e ective ranking functions. However, e ciency and e ectiveness are two competing forces and trading o e ectiveness for meeting e ciency constraints typical of production systems is one of the most urgent issues. This extended abstract shortly summarizes the work in [4] proposing CLEaVER, a new framework for optimizing LtR models based on ensembles of regression trees. We summarize the results of a comprehensive evaluation showing that CLEaVER is able to prune up to 80% of the trees and provides an e ciency speed-up up to 2:6x without a ecting the e ectiveness of the model.

T.Y.

Liu . Learning to rank for information retrieval . Foundations and Trends in IR . 2009 .

Lucchese ,

F. M.

Nardini ,

Orlando ,

Perego ,

Tonellotto , and

Venturini . QuickScorer: A fast algorithm to rank documents with additive ensembles of regression trees . In ACM SIGIR . 2015 .

Lucchese ,

F. M.

Nardini ,

Orlando ,

Perego ,

Tonellotto , and

Venturini . Exploiting CPU SIMD Extensions to Speed-up Document Scoring with Tree Ensembles . In ACM SIGIR . 2016 .

Lucchese ,

F. M.

Nardini ,

Orlando ,

Perego ,

Silvestri , S. Trani. PostLearning Optimization of Tree Ensembles for E cient Ranking . In ACM SIGIR . 2016 .

Capannini ,

Lucchese ,

F. M.

Nardini ,

Orlando ,

Perego ,

Tonellotto . Quality versus e ciency in document scoring with learning-to-rank models . In Information Processing & Management . 2016 .