The Setting
Finite-support approximation of an operator on a Lipschitz-free space asks for a common finite set of evaluation vectors for its outputs, with control of the error at prescribed anchors. Let be a pointed metric space satisfying unit separation: (1)
The real Lipschitz space is the normed linear space of real functions on satisfying (2) for some nonnegative Lipschitz bound , equipped with the Lipschitz norm: (2)✦
Evaluation at is the continuous linear functional on given by (3): (3)✦ The Lipschitz-free space is the Banach space given by the closed evaluation span (4) in ; the Dirac embedding regards these evaluations as elements of : (4)
The canonical dual pairing of with is evaluation of at , as in (5); here is finite and the coefficients are real: (5)✦
For , the Lipschitz point indicator is the function specified by (6): (6)✦ The coefficient functional is the canonical dual pairing against , as in (7): (7)✦ For a finite set , the finite-support projection is the bounded linear operator defined by (8): (8)
The Claim
Theorem 1. Suppose that for every , where . Let be real, let be a finite set of anchors, and let be a bounded finite-rank operator with and for every . There are a finite set and a bounded finite-rank operator satisfying (9): (9)✦
On the same bounded space, let be any bounded finite-rank operator and let . There is a finite set for which (10) holds: (10)✦
On an arbitrary pointed metric space satisfying unit separation, let be finite and let . The finite-support projection satisfies (11): (11)✦
In the paper, the result reads: for an integer and its constructed block , a bounded finite-rank operator on with and error less than at every anchor admits a finite set containing and a bounded finite-rank operator with range in , norm at most , and error less than at every anchor.
What Is Checked
The reduction and operator norm approximation are stated over bounded, unit-separated pointed metric spaces and bounded finite-rank operators on their closed evaluation spans; the assertion about a finite evaluation span is stated under unit separation alone. The evaluation functionals have norm at most , and the point indicators have Lipschitz norm at most .