断言失败乘以特征矩阵

时间:2013-11-23 07:43:46

标签: c++ codeblocks matrix-multiplication eigen least-squares

我正在编写一个c ++程序,用于插值中的最小二乘回归问题。我使用Eigen进行矩阵运算。我得到的问题是当我运行程序时它显示一个错误显示断言错误。这是我的代码:

#include <iostream>
#include <Eigen/Dense>
using Eigen::MatrixXd;
using namespace std;
int main()
{
    int i;
    int nmbrOfPoints;
    cout<<" Enter the number of data points : ";
    cin>>nmbrOfPoints;

    MatrixXd matY(nmbrOfPoints,1);       //initialize matrix Y
    MatrixXd matX(nmbrOfPoints,2);       //initialize matrix X
    MatrixXd matXdup(nmbrOfPoints,2);      //initialize matrix X duplicate
    MatrixXd matAns(2,1);


    for(i=0;i<nmbrOfPoints;i++)
    {
        matX(i,0)=1;                    // storing the 1 st column of the matrix x, all 1s.
        matXdup(i,0)=1;
    }

    cout<<"Enter all sample points (x and y values ): "<<endl;

    for(i=0;i<nmbrOfPoints;i++)
    {
        cin>>matX(i,1)>>matY(i,0); // read both (x,f(x)) ,, store x values to matrix x and y values to matrix y
    }

    for(i=0;i<nmbrOfPoints;i++)
    {
        matXdup(i,1)=matX(i,1);    //copying matrix x to its duplicate
    }

    cout<<"\n \n";
    cout << matX << endl;
    cout<<"\n \n";
    cout << matY << endl;
    cout<<"\n \n";
    cout << matXdup << endl;

    // find the transpose of matrix x

    cout << "\nHere is the transposed matrix x duplicate:\n" << endl;
    matXdup.transposeInPlace();


    cout << matXdup << endl;
    cout<<"\n \n";
    cout << matX << endl;

    //find the multiplication of x and transpose of x

    matX = matX* matXdup;   // now the matrix x holds the multiplication of transpose of x and x

    cout << "\nmultiplication of x and xdup:\n" << endl;
    cout << matX << endl;
    cout<<"\n \n";

    //find the inverse of x

    double q,a,b,c,d;

    a=matX(0,0);
    b=matX(0,1);
    c=matX(1,0);
    d=matX(1,1);

    q=1/((a*d)-(b*c));

    matX(0,0) = d*q;
    matX(0,1) = b*-1*q;             //now matrix x holds the inverse of x
    matX(1,0) = c*-1*q;
    matX(1,1) = a*q;

    cout<<"\n \n";
    cout << "\n inverse of x:\n" << endl;
    cout << matX << endl;

    //find the multiplication of transpose of x(x duplicate matrix) and y

     matY = matXdup* matY;   // now the matrix x duplicate holds the multiplication of y and x transpose

    //find the multiplication of x(inverse of xt*x) and matXdup (xt*y)

    // matAns = matY* matX;

     cout << "\nfinal answers :\n" << endl;
     cout << "\n *********************:\n" << endl;

     cout << matY << endl;
     cout<<"\n \n";
     cout << matX << endl;

     cout << "\nfinal answer FINAL :\n" << endl;
     cout << "\n *********************:\n" << endl;
     matAns = matY* matX;
     cout << matAns << endl;

     /*cout<<"\n matx dup = \n";
     cout << matXdup << endl;
     cout<<"\n maty =  \n";
     cout << matY << endl;
     cout<<"\n \n";*/

     return 0;


}

我从最终的乘法部分得到错误matAns = matY* matX

Assertion failed: a_lhs.cols() == a_rhs.rows() && "invalid matrix product" && "if you wanted a coeff-wise or a dot product use the respective explicit functions"

当我删除该语句代码有效时。到目前为止,代码工作正常。有人可以解释一下断言问题是什么以及如何解决它?

1 个答案:

答案 0 :(得分:2)

matY是2x1向量,matX是NxN矩阵,因此产品matY * matX无效。您确定不想将matX计算为:

matX = matXdup * matX;

和matAns as:

matAns = matX * matY;

BTW,无需明确转置matXduptransposeInPlace,您可以直接执行:

matX = matXdup.transpose() * matX;

此外,当在编译时知道维度并且此维度非常小时,更好地指定它。例如,matY应该是VectorXd。 matXdup.transpose() * matX的结果应该存储在Matrix2d对象中。然后调用inverse()而不是编写自己的反例程(您需要包含<Eigen/LU>

Matrix2d XX = matXdup.transpose() * matX; 
Vector2d Y = matXdup * matY;
Vector2d ans = XX.inverse() * Y;