Gradient boosting for convex cone predict and optimize problems
This work addresses the integration of prediction and optimization for decision-making in domains like operations research, though it appears incremental as it builds on existing SPO frameworks.
The paper tackled the problem of optimizing prediction models to minimize downstream decision regret by introducing dboost, the first general-purpose implementation of smart gradient boosting for predict-then-optimize problems, which reduces out-of-sample decision regret compared to state-of-the-art methods.
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.