For a real positive-definite symmetric matrix and vector , the quadratic functionhas unique global maximizer and maximum . Completing the square proves both claims:The subtracted quadratic form is nonnegative and vanishes only at . Solving is therefore sufficient, rather than merely a necessary first-order optimality condition. Numerically one solves the linear system instead of explicitly constructing the inverse.
For and , the maximizer is and its value is three. If were only positive semidefinite, uniqueness could fail, and a linear component along its kernel could make the objective unbounded above. This is why strict positive definiteness carries both existence and uniqueness.
Let be a symmetric positive-definite matrix, , and . Thenis the unique solution of minimizing the quadratic form under the affine optimization constraint . The minimum value is .
The denominator is positive because is a positive-definite matrix. The Lagrange multiplier equations give the displayed candidate and verify the constraint. For a global proof, any feasible perturbation satisfies . HenceThis is positive for every nonzero . The result also handles several independent quadratic blocks by taking to be their block diagonal matrix and concatenating their variables.
A real positive-definite symmetric matrix defines an inner product and norm . For a nonzero direction , every vector has the unique decompositionSubstitution proves the orthogonality and gives the corresponding Pythagorean identity . Thus is the closest vector to on the line spanned by , in this metric.
For , and , one gets , unlike the Euclidean projection coefficient . The geometry weights coordinate errors differently. Under the invertible change of variables , it becomes ordinary Euclidean orthogonal decomposition. This is useful for separating a constrained direction from free directions in quadratic optimization.
Articles by others on the same topic
There are currently no matching articles.