CVX求解线性规划（LP）、二阶锥规划（SOC）、半定规划 (SDP)问题

「已注销」

已于 2023-08-22 17:04:23 修改

阅读量1.8k

点赞数

文章标签： matlab 数学建模开发语言

于 2023-06-07 10:49:53 首次发布

本文链接：https://blog.csdn.net/asdzdf/article/details/131083020

版权

CVX toolbox

CVX是一个常用的matlab凸优化求解器，但是不能求解非凸优化问题，而且优化问题必须满足CVX的DCP规则，CVX提供了使用手册CVX user guide。

下面分别展示用CVX求解LP,SOCP和SDP问题。

LP

优化问题：
在这里插入图片描述
代码：

%%
clc,clear;
close all
%随机初始化参数
s=rng(1)
m = 20; n = 10; p = 4;
A = randn(m,n); b = randn(m,1);
C = randn(p,n); d = randn(p,1); e = rand;
cvx_begin
    variable x(n) %声明变量
    minimize( norm( A * x - b, 2 ) ) %目标函数
    subject to
        C * x == d %等式约束
        norm( x, Inf ) <= e %不等式约束
cvx_end
输出结果

输出结果：

Calling SDPT3 4.0: 56 variables, 22 equality constraints
   For improved efficiency, SDPT3 is solving the dual problem.
------------------------------------------------------------

 num. of constraints = 22
 dim. of socp   var  = 41,   num. of socp blk  = 11
 dim. of linear var  = 11
 dim. of free   var  =  4 *** convert ublk to lblk
*******************************************************************
   SDPT3: Infeasible path-following algorithms
*******************************************************************
 version  predcorr  gam  expon  scale_data
    NT      1      0.000   1        0    
it pstep dstep pinfeas dinfeas  gap      prim-obj      dual-obj    cputime
-------------------------------------------------------------------
 0|0.000|0.000|1.4e+01|8.5e+00|3.0e+03| 8.795268e+00  0.000000e+00| 0:0:00| chol  1  1 
 1|0.922|0.470|1.1e+00|4.5e+00|7.3e+02| 1.475473e+01 -4.178097e+00| 0:0:00| chol  1  1 
 2|1.000|0.966|2.0e-05|1.6e-01|3.9e+01| 1.554422e+01 -4.431746e+00| 0:0:00| chol  1  1 
 3|0.977|0.870|2.2e-06|2.2e-02|6.2e+00| 8.064512e-01 -4.453998e+00| 0:0:00| chol  1  1 
 4|0.758|0.662|6.1e-07|7.4e-03|2.1e+00|-2.191328e+00 -4.153897e+00| 0:0:00| chol  1  1 
 5|1.000|0.230|3.2e-08|5.7e-03|7.7e-01|-3.442874e+00 -4.106998e+00| 0:0:00| chol  1  1 
 6|1.000|0.502|1.4e-08|2.8e-03|2.6e-01|-3.806851e+00 -4.030127e+00| 0:0:00| chol  1  1 
 7|0.971|0.646|3.2e-09|1.0e-03|6.5e-02|-3.942257e+00 -3.996587e+00| 0:0:00| chol  1  1 
 8|1.000|0.498|1.3e-09|5.0e-04|2.1e-02|-3.972640e+00 -3.988766e+00| 0:0:00| chol  1  1 
 9|0.996|0.927|3.5e-10|3.7e-05|9.7e-04|-3.982939e+00 -3.983604e+00| 0:0:00| chol  1  1 
10|0.954|0.985|2.2e-10|3.3e-06|5.0e-05|-3.983425e+00 -3.983451e+00| 0:0:00| chol  1  1 
11|0.944|0.979|2.4e-10|1.7e-07|3.2e-06|-3.983440e+00 -3.983442e+00| 0:0:00| chol  1  1 
12|0.969|0.977|1.9e-09|1.1e-08|2.1e-07|-3.983442e+00 -3.983442e+00| 0:0:00| chol  1  2 
13|1.000|0.984|6.1e-10|7.5e-10|8.6e-09|-3.983442e+00 -3.983442e+00| 0:0:00|
  stop: max(relative gap, infeasibilities) < 1.49e-08
-------------------------------------------------------------------
 number of iterations   = 13
 primal objective value = -3.98344194e+00
 dual   objective value = -3.98344194e+00
 gap := trace(XZ)       = 8.59e-09
 relative gap           = 9.58e-10
 actual relative gap    = 4.35e-10
 rel. primal infeas (scaled problem)   = 6.15e-10
 rel. dual     "        "       "      = 7.47e-10
 rel. primal infeas (unscaled problem) = 0.00e+00
 rel. dual     "        "       "      = 0.00e+00
 norm(X), norm(y), norm(Z) = 3.0e+00, 4.2e+00, 5.9e+00
 norm(A), norm(b), norm(C) = 1.8e+01, 2.0e+00, 6.5e+00
 Total CPU time (secs)  = 0.16  
 CPU time per iteration = 0.01  
 termination code       =  0
 DIMACS: 6.1e-10  0.0e+00  1.3e-09  0.0e+00  4.4e-10  9.6e-10
-------------------------------------------------------------------
 
------------------------------------------------------------
Status: Solved
Optimal value (cvx_optval): +3.98344

SOCP

在这里插入图片描述
代码：

clc,clear;
close all;
s = rng(1);
y0 = rand(1);
N = 2;%q的维度
d = rand(N,1);
cvx_begin
variable q(N,1)
variable y nonnegative
maximize (log(1+1/y0)-(y-y0)/(y0*(y0+1)))
subject to
    norm([2*(q-d);1-y])<=1+y %SOC约束

cvx_end

输出结果：

Calling SDPT3 4.0: 6 variables, 2 equality constraints
------------------------------------------------------------

 num. of constraints =  2
 dim. of socp   var  =  4,   num. of socp blk  =  1
 dim. of linear var  =  2
*******************************************************************
   SDPT3: Infeasible path-following algorithms
*******************************************************************
 version  predcorr  gam  expon  scale_data
    NT      1      0.000   1        0    
it pstep dstep pinfeas dinfeas  gap      prim-obj      dual-obj    cputime
-------------------------------------------------------------------
 0|0.000|0.000|3.8e+00|4.9e+00|2.0e+02| 1.692250e+01  0.000000e+00| 0:0:00| chol  1  1 
 1|0.697|0.741|1.1e+00|1.3e+00|9.0e+01| 1.371567e+01 -2.019835e+01| 0:0:00| chol  1  1 
 2|1.000|1.000|2.0e-05|6.4e-03|1.3e+01| 5.376201e+00 -7.856240e+00| 0:0:00| chol  1  1 
 3|0.935|0.984|5.5e-06|7.4e-04|4.8e-01| 3.627761e-01 -1.139014e-01| 0:0:00| chol  1  1 
 4|0.706|1.000|1.6e-06|6.6e-05|1.8e-01| 1.698983e-01 -6.243616e-03| 0:0:00| chol  1  1 
 5|0.986|0.986|2.3e-08|7.6e-06|2.5e-03| 2.408733e-03 -7.809152e-05| 0:0:00| chol  1  1 
 6|0.989|0.989|3.0e-10|7.2e-07|2.8e-05| 2.656438e-05  1.278095e-07| 0:0:00| chol  1  1 
 7|0.989|0.989|3.0e-12|8.0e-09|3.0e-07| 2.919599e-07  1.399982e-09| 0:0:00| chol  2  3 
 8|0.989|1.000|3.6e-10|1.0e-12|5.0e-09| 4.457611e-09 -2.774934e-10| 0:0:00|
  stop: max(relative gap, infeasibilities) < 1.49e-08
-------------------------------------------------------------------
 number of iterations   =  8
 primal objective value =  4.45761069e-09
 dual   objective value = -2.77493445e-10
 gap := trace(XZ)       = 4.96e-09
 relative gap           = 4.96e-09
 actual relative gap    = 4.74e-09
 rel. primal infeas (scaled problem)   = 3.59e-10
 rel. dual     "        "       "      = 1.00e-12
 rel. primal infeas (unscaled problem) = 0.00e+00
 rel. dual     "        "       "      = 0.00e+00
 norm(X), norm(y), norm(Z) = 1.4e+00, 2.8e-01, 1.3e+00
 norm(A), norm(b), norm(C) = 3.2e+00, 2.4e+00, 2.7e+00
 Total CPU time (secs)  = 0.18  
 CPU time per iteration = 0.02  
 termination code       =  0
 DIMACS: 4.3e-10  0.0e+00  1.0e-12  0.0e+00  4.7e-09  5.0e-09
-------------------------------------------------------------------
 
------------------------------------------------------------
Status: Solved
Optimal value (cvx_optval): +1.92888

SDP

clc,clear;
close all;

N = 10;
A = rand(N) +1i*rand(N);
A = (A*A');

cvx_begin
variable X(N,N) semidefinite  hermitian
minimize (real(trace(A*X)))
subject to
    real(trace(X))==1
    
cvx_end

码字不易，欢迎点赞、关注、转发、收藏。后续会持续分享相关内容。

知乎专栏，CVX求解线性规划（LP）、二阶锥规划（SOC）

「已注销」

关注

0
点赞
踩
14

收藏

觉得还不错? 一键收藏
2
评论
CVX求解线性规划（LP）、二阶锥规划（SOC）、半定规划 (SDP)问题

CVX是一个常用的matlab凸优化求解器，但是不能求解非凸优化问题，而且优化问题必须满足CVX的DCP规则，CVX提供了使用手册CVX user guide。下面分别展示用CVX求解LP,SOCP和SDP问题。
复制链接

扫一扫