The Setting
The ramp construction concerns normalized differences of real functions on a finite vertex set, measured by a weighted energy on edges. Let and be finite vertex and edge sets, with endpoint maps . Assume , every vertex is incident to an edge, and every pair of distinct vertices is joined by an edge in either orientation. Give each edge a scale and a weight , with ; write for the edge probability measure.
The incidence matrix and the normalized gradient are the matrices in (1). The normalized-gradient space is a subspace of . (1) A retraction onto is a matrix that fixes and factors through ; its row norm is denoted by , as in (2). (2)
For scalar thresholds and , the tail truncation is the map that retains a scalar when and otherwise vanishes; the band truncation retains it when and otherwise vanishes.✦ The dyadic ramp , for an integer , is the sum of band truncations in (3).✦ (3) The vector tail truncation on is the map that applies the scalar tail truncation to each coordinate. For , the power intensity is the measure on the positive half-line with density ✦, and the Pareto law is its restriction to : (4) Let be a nonempty finite set, put , and choose vectors with for every edge . The uniform mixture of measures on a common measurable space is the measure given by their arithmetic average; the Pareto vector mixture is the measure on in (5), formed from and the law in (4).✦ The symbol denotes pushforward. (5) The coordinatewise ramp is the map that applies the scalar map in (3) on each edge, and is the scalar -energy under . For , the ramp variance constant and the graph parameter are the numbers in (6). (6)
The Claim
Theorem 1. For , a nonempty finite family normalized by on every coordinate, and , the mixture in (5) satisfies the moment identity (7): (7)✦ For every integer , the even-length scalar ramp has the lower bound (8): (8)✦
Suppose . There is such that the following projection and retraction estimates hold for every integer . For the projection estimate, let be a probability measure on , let be a real matrix on , and give nonnegative weights of sum one. Assume , , for every , and for -almost every . Assume also that is integrable for every and , with the second-moment bounds in (9): (9) Then the projection error obeys (10): (10)✦
For the graph in (1) and the normalized family , with the positive weights and pairwise adjacency specified above, there is a retraction satisfying (11), with and the constants from (5) and (6): (11)✦
For each there is , independent of the graph, its weights and the family , such that a probability measure on satisfies the bounded gradient jump estimates (12): (12)✦
In the paper, the result reads: with and scale parameter , for a vector of law and an integer , the coordinate truncations give and , every edge has -energy of order , the weighted second moment of is at most and each coordinate second moment is at most , while and , giving relative error at most and the choice .
What Is Checked
The moment and scalar energy assertions concern the truncations and Pareto measures in (3) and (5); the projection assertion concerns a probability law fixed almost surely by a matrix with the tail bounds (9), and the retraction and jump assertions concern finite normalized-gradient spaces with positive edge scales and normalized positive weights, with every vertex incident to an edge and every pair of distinct vertices adjacent.