Gradient boosting for convex cone predict and optimize problems

DOI10.1016/J.ORL.2022.12.006MaRDI QIDQ6161287zbMATH OpenOpenAlexFDO

Authors Roy H. Kwon

Publication date 27 June 2023

Published in Operations Research Letters (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2204.06895

zbMATH Keywords

convex optimizationdecision tree regressiongradient boosting

Mathematics Subject Classification ID

Semidefinite programming (90C22) Convex programming (90C25)

Abstract: Prediction models are typically optimized independently from decision optimization. A smart predict then optimize (SPO) framework optimizes prediction models to minimize downstream decision regret. In this paper we present dboost, the first general purpose implementation of smart gradient boosting for `predict, then optimize' problems. The framework supports convex quadratic cone programming and gradient boosting is performed by implicit differentiation of a custom fixed-point mapping. Experiments comparing with state-of-the-art SPO methods show that dboost can further reduce out-of-sample decision regret.

Recommendations

• Mixed-Integer Convex Nonlinear Optimization with Gradient-Boosted Trees Embedded
• Accelerated gradient boosting
• Optimization of tree ensembles
• Predictive stochastic programming
• Predictive online convex optimization

Cites work

• Greedy function approximation: A gradient boosting machine.
• Hybrid classification and reasoning for image-based constraint solving
• Implicit Functions and Solution Mappings
• On the Numerical Solution of Heat Conduction Problems in Two and Three Space Variables
• Random forests

Cited in

(3)

• Greedy function approximation: A gradient boosting machine.
• Mixed-Integer Convex Nonlinear Optimization with Gradient-Boosted Trees Embedded
• An improved feature-based forecast combination method using rolling origin evaluation

This page was built for publication: Gradient boosting for convex cone predict and optimize problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6161287)