机器学习能否预测困难组合优化问题的解值?
人工智能
2020-03-09 v1 机器学习
摘要
我们考察机器学习能否根据输入预测困难组合优化问题的最终目标函数值。我们的背景是模式削减问题,它是下料问题中具有工业重要性但困难的一个方面。机器学习似乎比朴素模型具有更高的预测精度,将平均绝对百分比误差(MAPE)从 12.0% 降至 8.7%。
引用
@article{arxiv.2003.03181,
title = {Can ML predict the solution value for a difficult combinatorial problem?},
author = {Constantine Goulimis and Gastón Simone},
journal= {arXiv preprint arXiv:2003.03181},
year = {2020}
}
备注
The work will be presented in the PMO104 workshop (2 & 3 April 2020) organised by the German Operations Research Society in Bad Honnef