Convex Hull Formulations for Mixed-Integer Multilinear Functions

07/29/2018

∙

by Harsha Nagarajan, et al.

∙

Los Alamos National Laboratory

∙

In this paper, we present convex hull formulations for a mixed-integer, multilinear term/function (MIMF) that features products of multiple continuous and binary variables. We develop two equivalent convex relaxations of an MIMF and study their polyhedral properties in their corresponding higher-dimensional spaces. We numerically observe that the proposed formulations consistently perform better than state-of-the-art relaxation approaches.

READ FULL TEXT

Convex Hull Formulations for Mixed-Integer Multilinear Functions

Piecewise Polyhedral Formulations for a Multilinear Term

Partition-based formulations for mixed-integer optimization of trained ReLU neural networks

Gaining or Losing Perspective for Piecewise-Linear Under-Estimators of Convex Univariate Functions

Tightness of prescriptive tree-based mixed-integer optimization formulations

The CCP Selector: Scalable Algorithms for Sparse Ridge Regression from Chance-Constrained Programming

Outlier detection in regression: conic quadratic formulations

Convex Relaxations of Bregman Divergence Clustering

Convex Hull Formulations for Mixed-Integer Multilinear Functions

Related Research

Piecewise Polyhedral Formulations for a Multilinear Term

Partition-based formulations for mixed-integer optimization of trained ReLU neural networks

Gaining or Losing Perspective for Piecewise-Linear Under-Estimators of Convex Univariate Functions

Tightness of prescriptive tree-based mixed-integer optimization formulations

The CCP Selector: Scalable Algorithms for Sparse Ridge Regression from Chance-Constrained Programming

Outlier detection in regression: conic quadratic formulations

Convex Relaxations of Bregman Divergence Clustering