Various Eigenvector Results in C Versus Python

Question

So, I noticed that I have different answers to the eigendecomposition of the 4x4 matrix of all 1s.

In Python using numpy.linalg.eig:

matrix = numpy.ones((M,M), dtype=float);
values, vectors = numpy.linalg.eig(matrix);

Python result:

V1: [-0.866025 +0.288675 +0.288675 +0.288675]
V2: [+0.500000 +0.500000 +0.500000 +0.500000]
V3: [+0.391955 +0.597433 -0.494694 -0.494694]
V4: [+0.866025 -0.288675 -0.288675 -0.288675]

In C Using LAPACK DSYEV:

#define NN 4
#define LDA NN
void main(){
    int n = NN, lda = LDA, lwork=NN*NN*NN*NN*NN, info;
    char both = 'V';
    char uplo = 'U';
    double w[NN*NN];
    double work[NN*NN*NN*NN*NN];
    double a[LDA*NN] = {
       1,  1,  1,  1,
       1,  1,  1,  1,
       1,  1,  1,  1,
       1,  1,  1,  1 
    };

    dsyev_(&both, &uplo, &n, a, &lda, w, work, &lwork, &info);
    return;
}

C DSYEV Result:

V1: +0.000596 +0.000596 -0.707702 +0.706510 
V2: +0.500000 +0.500000 -0.499157 -0.500842 
V3: +0.707107 -0.707107 -0.000000 +0.000000 
V4: +0.500000 +0.500000 +0.500000 +0.500000  

In C Using LAPACK DGEEV:

#define NN 4
#define LDA NN
#define LDVL NN
#define LDVR NN
void main() {
    char compute_left = 'V';
    char compute_right = 'V';
    int n = NN, lda = LDA, ldvl = LDVL, ldvr = LDVR, info, lwork=2*NN*NN;
    double work[2*NN*NN];
    double wr[NN], wi[NN], vl[LDVL*NN], vr[LDVR*NN];
    double a[LDA*NN] = {
       1,  1,  1,  1,
       1,  1,  1,  1,
       1,  1,  1,  1,
       1,  1,  1,  1 
    };
    dgeev_( &compute_left, &compute_right, &n, a, &lda, wr, wi, vl, &ldvl, vr, &ldvr, work, &lwork, &info );
    return;
}

C DGEEV Result:

V1: -0.866025 +0.288675 +0.288675 +0.288675 
V2: -0.500000 -0.500000 -0.500000 -0.500000 
V3: -0.000000 -0.816497 +0.408248 +0.408248 
V4: -0.000000 -0.000000 -0.707107 +0.707107 

The results are all different!

So, I have two main questions:

• Why? Is it related to degeneration in 1 matrix?
• How can I replicate Python In C results?

Any insight would be appreciated.

Answer 1

. : 4 0. 4 - , [1,1,1,1], . 0 3- x_1 + x_2 + x_3 + x_4 = 0. - numpy, , , .

, DGEEV , , 0- .

Answer 2

: 0 ( values Python script). (), , , , , . , , , , , . , , 0.

4 ( ), , ( ) . , .

, , . , , Wolfram Alpha.

Update

, a M by M M . ( ) , , ( v1 v2), v3 = a*v1 + b*v2, a b - .