non-zero values constraint in QP

[Pravin Bezwada]

Hi all,

I have QP problem as below:

def getweights(params, cov):

    (m,n) = np.shape(params)

    P = matrix(cov)

    q = matrix([0.] * m, (m, 1))

    A = matrix(params.T)

    b = matrix([0.] * n, (n, 1))

    res = solvers.qp(P=P, q=q, A=A, b=b)    

    return res['x']

This always returns zero for all x. I want to add a non zero condition to each x. Could someone please help me?

Best regards,

Pravin

[Martin]

It looks like you're solving a QP with equality constraints only. This problem can be solved analytically by solving the equation

[PAAT0][xλ]=[−qb]

and since q=0 and b=0 in your case, it is clear that x=0 is a solution. Note that the constraint that x must be nonzero is a nonconvex constraint, but you can add the constraint that x is nonnegative, but this clearly does not help in this case.

[Martin]

The equations did not show up as expected, so here there are in text:

P*x + A.T*lambda = -q

A*x = b