经典的SDR算法：求解二次优化问题，附加高斯随机化确保所求的解的秩为1。附加代码

1、原始问题

2、先展开，再去掉常数项

3、改成最小化 优化问题（取反）

4、引入辅助变量，变成高一维


5、放松约束

6、恢复问题解

本文以武庆庆老师的Intelligent Reflecting Surface Enhanced Wireless
Network: Joint Active and Passive Beamforming
Design这篇文章为例来求解二次优化问题。

    hd_x = h_d_all(:,nnnum_subset);
    G_x = G_all(:,:,nnnum_subset);
    hd_x_H = hd_x';
    
    
    Phi=diag(hr_H)*G_x;
    R_=[Phi*Phi' Phi*hd_x;hd_x_H*Phi' 0];
    
    r=round(log10(mean(mean(R_))));
    R=R_;
    ishermitian(R);
    
    
    cvx_begin
    variable V(N+1,N+1) symmetric
    maximize(trace(R*V))
    subject to
    for n=1:N+1
        V(n,n)==1;
    end
    V == semidefinite(N+1);
    cvx_end
    
    
    [U,W,Z] = svd(V);%U==Z
    obj_max=-100;
    n=length(W);
    for j=1:1000
        
        r=sqrt(1/2)*(randn(n,1)+1i*randn(n,1));%sqrt(var/2)*(randn(1,N)+1i*randn(1,N))
        v_=U*sqrt(W)*r;
        obj=(v_'*R*v_);
        
        if obj>=obj_max
            obj_max=obj;
            v_max=v_;
        end
        
        
    end
    
    
    
    t=v_max(N+1);
    for j=1:N
        
        v(j,:)=exp(1i*angle(v_max(j)/t));
        %v(j,:)=v_max(j)/t;
    end