You should set the (2,3) and (3,2) elements to exactly zero.

I am asking where these near zero input data come from? Are the (3,1) and (1,3) elements meaningful numbers which happen to be very small magnitude?

You should set the (2,3) and (3,2) elements to exactly zero.

I am asking where these near zero input data come from? Are the (3,1) and (1,3) elements meaningful numbers which happen to be very small magnitude?

Yes I set the (2,3) and (3,2) elements to exactly zero.

and (2,1), (3,1), (1,2), (1,3) elements are meaningful numbers

Each element is obtained by the following equation

As a result of the calculation, they have very small values

Try changing the units. But perhaps the problem is inherently badly scaled. Are you using Mosek 10.x?

It may be that inaccurate/solved is as good as you can get it, in which case, the problem has been inaccurately solved.

Thank you for your reply

My MOSEK Version is 10.1.17 and I got CVX Status is Solved but

MOSEK warning 710 (MSK_RES_WRN_ZEROS_IN_SPARSE_COL): #1 (nearly) zero elements are specified in sparse col ‘’ (123) of matrix ‘A’.

I wonder if changing units means multiplying matrix A by scale, My objective function is x^T * A * x.

Simply applying scaling factors such as 10^3 or 10^6 to matrix A still results in the same warning, and sometimes even yields an ‘unbounded’ status.

Perhaps you can show reproducible code, preferably with all inputs, as well as all CVX and Mosek output. That will allow forum readers from Mosek to make a better assessment of the situation.

The warning about matrix A refers to the matrix A of linear constraints. Mosek refers to the input it receives, it does not know anything about what you have in your cvx model.

If you want you can make Mosek write the problem to a readable file format, see how the problem looks and then work out where the problematic data is in your model. This is usually not obvious so be patient.