简体   繁体   English

如何使用神经网络进行多目标优化?

[英]How to do multi-objective optimization using neural networks?

I have five decision variables, each having a specific range. 我有五个决策变量,每个都有一个特定的范围。 I need to find a combination of these variables so as to maximize one of my objectives while minimizing the other at the same time. 我需要找到这些变量的组合,以便最大化我的一个目标,同时最小化另一个目标。 I have prepared a datasheet of randomly generated variables with respective values of the 2 objective functions. 我准备了一个随机生成的变量的数据表,其中包含2个目标函数的相应值。 Please suggest me how to approach the solution using neural networks. 请建议我如何使用神经网络解决方案。

My objective function involves thermodynamic calculations. 我的目标函数涉及热力学计算。 If interested you can have a look at the objective functions here : 如果有兴趣,可以在这里查看目标函数:

There are many methods, but the easiest way is "linear scalarization". 有很多方法,但是最简单的方法是“线性标量”。
You can add objectives to make single objective. 您可以添加目标以创建单个目标。
While doing this, you can weight objectives considering priority. 在执行此操作时,您可以权衡目标的优先级。
(Making linear combination of multiple objectives) (使多个目标线性组合)

See examples: 查看示例:
Variational AE loss (regularization loss + reconstruction loss) 可变AE损失(正则化损失+重建损失)
Associative DA loss (classification loss + walker loss + visit loss) 关联DA损失(分类损失+步行者损失+访问损失)

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM