The Setting
Let be a finite vertex set, write , and let be a nonempty finite edge set with endpoint maps and positive scales . The incidence matrix records the difference of endpoint indicators✦, and the normalized gradient is obtained by dividing row of by , as in (1).✦ The normalized-gradient space is the range of . (1)
For a real matrix , a retraction onto its range is a matrix that fixes that range and has all its values in it; the set of these retractions is given by (2). (2)✦
The row norm of is the sum of the absolute values of its entries in row .✦ For a real inner product space and data , the matrix action is the finite linear combination in (3), which also specifies . (3)
Weighted Projections
Choose an orthonormal eigenvector matrix of the Gram matrix , with corresponding eigenvalues , . The Gram eigenbasis coordinates and the orthogonal range projection are the matrices in (4); here for and . (4)
For real row weights , let , and let be the diagonal matrix with reciprocal diagonal entries and the same convention at zero. The weighted orthogonal projection is the matrix defined by (5). (5)✦ For strictly positive , write for with squared norm .
The Claim
Theorem 1 (Weighted projection and controlled retraction). For any real matrix , strictly positive weights , and , the weighted projection satisfies (6). (6)✦
Suppose every vertex is incident to an edge and every two distinct vertices are joined by an edge. If is a strictly positive probability on , the normalized gradient satisfies the weighted row estimate (7). (7)✦
Under these endpoint assumptions, suppose also , let be a strictly positive probability on , and let be finite. For data , , with for every , there is a matrix satisfying (8). (8)✦
Under the same assumptions on the endpoints, , and , let , , be positive bounds and let the data satisfy for every . Then there is a matrix satisfying (9). (9)✦
In the paper, the result reads: for every strictly positive probability on , there is a probability with such that the orthogonal projection onto in satisfies , hence , and, with denoting the identity on , both and are contractions on .
What Is Checked
The energy estimate is stated for arbitrary finite real matrices and positive weights; the remaining statements concern normalized gradients with the indicated endpoint conditions, and the retraction statements allow finite families of data in an arbitrary real inner product space.
For every real matrix , its Gram matrix is symmetric.✦