NNLS问题与某些零数量作为约束

发布于 2025-02-07 12:14:08 字数 139 浏览 2 评论 0原文

我正在做一个优化问题,以最大程度地减少AX和B的正方形差异。

a是m*n的矩阵,其中x是非负n载体。我想添加一个约束,即从向量X中的n个元素中的p分别为零。

我应该如何制定条件?听起来有NNLS优化问题的NCP集。

谢谢

I am doing an optimization problem To minimise square difference of Ax and b.

A is a matrix of m*n where x is non negative n vector. I would like to add a constraint that p out of n elements in vector x is zero.

How should I formulate the condition? It sounds to have nCp set of nnls optimization problem.

Thanks

如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

扫码二维码加入Web技术交流群

发布评论

需要 登录 才能够评论, 你可以免费 注册 一个本站的账号。

评论(1

素手挽清风 2025-02-14 12:14:08

标准公式是添加基数约束

 min ||Ax-b||^2
 x[i] ≤ δ[i]*U
 sum(i,δ[i]) = n-p
 δ[i] ∈ {0,1}
 x[i] ∈ [0,U]  

这是MIQP(混合二次二次编程)模型。 ux [i]上的上行(应该不太大)。 Δ[i]二进制决策变量,含义:

δ[i] = 1:  variable x[i] is allowed to vary between 0 and U
δ[i] = 0:  variable x[i] must be zero  

有些冗余:模型同时计算xΔ 。 (这就是“变量”的含义。)

MIQP求解器很容易获得。

A standard formulation for this is to add a cardinality constraint:

 min ||Ax-b||^2
 x[i] ≤ δ[i]*U
 sum(i,δ[i]) = n-p
 δ[i] ∈ {0,1}
 x[i] ∈ [0,U]  

This is an MIQP (Mixed-Integer Quadratic Programming) model. U is an upperbound on x[i] (should be not too large). δ[i] is a binary decision variable with meaning:

δ[i] = 1:  variable x[i] is allowed to vary between 0 and U
δ[i] = 0:  variable x[i] must be zero  

Somewhat redundantly: the model computes both x and δ. (That is what "variable" means.)

MIQP solvers are readily available.

~没有更多了~
我们使用 Cookies 和其他技术来定制您的体验包括您的登录状态等。通过阅读我们的 隐私政策 了解更多相关信息。 单击 接受 或继续使用网站,即表示您同意使用 Cookies 和您的相关数据。
原文