Generalization Bounds of Surrogate Policies for Combinatorial Optimization Problems

TL;DR

This paper develops a theoretical framework for analyzing the generalization of smoothed policies in combinatorial optimization, introducing bounds that decompose excess risk into bias, estimation, and optimization errors.

Contribution

It introduces a novel geometric analysis of perturbation bias via the fan-crossing probability and provides generalization bounds for smoothed policies in combinatorial optimization.

Findings

Perturbation bias is controlled by the fan-crossing probability.

The statistical estimation error scales as 1/(λ√n).

Kernel Sum-of-Squares methods help mitigate optimization curse of dimensionality.

Abstract

Many real-world decision problems require solving, again and again, combinatorial optimization instances drawn from a common distribution. A recent line of structured learning methods exploits this regularity by learning policies that pair a statistical model with a tractable combinatorial oracle, instead of solving each instance independently. Training such policies is notoriously difficult, however: the resulting empirical risk is piecewise constant in the model parameters, which hinders gradient-based optimization, and only a few theoretical guarantees have been provided so far. We address this issue by analyzing smoothed (perturbed) policies: adding controlled random perturbations to the direction used by the linear oracle yields a differentiable surrogate risk and improves generalization. Our main contribution is a generalization bound that decomposes the excess risk into $(\mathit{i})$ perturbation bias, $(\mathit{ii})$ statistical estimation error, and $(\mathit{iii})$ optimization error. The perturbation bias is controlled by the \emph{fan-crossing probability}, a new geometric quantity measuring the likelihood that a perturbation changes the oracle solution. We introduce two complementary conditions to bound it--the \emph{Uniformly Bounded Density} (UBD) property, yielding a sharp ${O}(\lambda)$ bias, and the weaker \emph{Uniform Weak moment} (UW) property, yielding a sub-linear bound--both capturing the geometric interaction between the statistical model and the normal fan of the feasible polytope. The statistical estimation error is controlled via a uniform deviation bound over the policy class, with rate ${O}(1/(\lambda\sqrt{n}))$ that scales inversely in the smoothing parameter. Concerning the optimization error, we exploit kernel Sum-of-Squares methods to mitigate the curse of dimensionality of global optimization.