1 | initial version |
There is a simpler explanation, which is that the quadratic form
[ 2 -1 ]
[ * 2 ]
Maps to the matrix
[ 2 -1 ]
[ 0 2 ]
Which is equivalent to the symmetric matrix
[2 -1/2]
[-1/2 2]
Which can be gotten with A.Gram_matrix().
The A.matrix() function listed in the answer above is the Hessian, which is double the Gram matrix.
The documentation for Gram matrix gives as an example a Quadratic form with diagonal matrix. Should probably be updated to include off-diagonal terms to make this clearer.