Uma formulação para a máquina de aprendizagem mínima baseada em programação linear

Abstract

Minimal Learning Machine (MLM) is a supervised learning method whose basic principle is based on a linear mapping between distances in the input and output spaces, followed by an optimization process to, based on estimated distances, provide an estimate for the output in a typical regression case. The MLM test step involves solving a non-convex optimization problem and it may su er from local minima problems. In this paper, we present a formulation for the out-of-sample step using linear programming.The experiments show that the proposed method achieves similar performance to that obtained with the original algorithm, additionally producing results with small variance.