NLP模型需要很长时间才能完成 [英] NLP model takes a very long to complete

问题描述

我有一个NLP 模型，有202个决策（101个实数[0，无穷大]，101个整数[0,1]），一个目标和85个以上的约束。在我的工作站（Xeon E5420（2.5GHz），3GB内存）上，使用Microsoft.SolverFoundation.Solvers.HybridLocalSearchSolver运行需要数小时。
我需要第三方解决方案来解决这个问题吗？或者NLP模型无法解决？

I have a NLP model that has 202 decisions (101 reals[0, Infinity], 101 integers[0, 1]), one goal and 85+ constraints. On my workstation (Xeon E5420 (2.5GHz), 3GB of ram), it takes hours to run using Microsoft.SolverFoundation.Solvers.HybridLocalSearchSolver. Do I need a third party solver to solve this problem? Or is NLP model unsolvable?

请建议

John Yung

John Yung

推荐答案

如果你能够重写模型以便唯一的约束是变量的界限，那么NelderMeadSolver可能会运行正常。否则，我会考虑使用第三方解算器，如Frontline或Ziena's。

If you are able to rewrite the model so that the only constraints are bounds on the variables, NelderMeadSolver will probably work out okay. Otherwise, I would consider using a third party solver such as Frontline or Ziena's.

Nate

这篇关于NLP模型需要很长时间才能完成的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！