A reformulation to Embedding a Neural Network in a linear program without integer variables

Abstract

In this technical report, a new formulation for embedding a neural network into an optimization model is described. This formulation does not require binary variables to properly compute the output of the neural network for specific types of problems. Preliminary experiments show that this reformulation resulted in faster computation times when solving a proposed showcase model, in which non-linearity is necessary to be computed. This is in comparison with the classic formulation and off-the-shelf tools of commercial solvers.