The Setting
For a finite subset of a real space, the dimension problem asks for a realization in finitely many real coordinates with prescribed distortion. Here . The constructions concern coordinate moments of fixed sums of symmetric random vectors and a cutoff that discards a whole vector.
The coordinate distance between is the distance defined in (1). (1) For a family size and a distortion , an admissible dimension for is a nonnegative integer such that every injective family in , over any measure space, admits and a scale satisfying (2) for every . (2) The universal coordinate dimension is the infimum in the nonnegative integers of the admissible dimensions, as in (3). (3)✦
Gradient Laws and Ramp Parameters
Let and be finite vertex and edge sets, with endpoint maps and real edge scales . The incidence matrix is the matrix with entries in (4); the normalized gradient is its product with the indicated diagonal matrix, and its image is the normalized-gradient space. (4)
A symmetric probability law on is a probability measure invariant under . For a measure on and , the fixed convolution power is the measure defined by (5), where denotes unit mass at zero. (5) For a measure on , an exponent , coordinate weights , and a target , write for its coordinate th moment and for the weighted error from , as in (6). (6) The global cutoff at a real threshold is the map defined in (7). (7)✦
The power intensity is the measure on the positive half-line with density , and the Pareto measure is its restriction to , as in (8). (8) For real , the band truncation retains a scalar precisely when its magnitude belongs to .✦ The dyadic ramp sum , for , is the sum of these truncations in (9). (9) The ramp variance constant , the target moment , and the scales for are defined by (10). (10)✦
The Claim
Theorem 1. Let , , , and . Suppose is a symmetric probability law on , , and . Assume for every , almost surely, and the moment conditions in (11). (11) The weighted moment estimate (12) holds with a constant depending only on . (12)✦
Let be a symmetric probability law on , , and . Suppose for all , almost surely, and for each . For every and with , the exponential integrability and bound are (13). (13)✦
Take , a probability law on , and . Suppose every is integrable and the exponential integrability and bound in (13) hold for in place of . With a constant depending only on , the cutoff error satisfies (14) for every . (14)✦
Fix . Suppose is nonempty, , every vertex meets an edge, and every two distinct vertices are endpoints of some edge. Take with , and a finite nonempty family in satisfying for every . For every and , there is a probability law on satisfying (15), with depending only on . (15)✦
Suppose , , and . For each integer , parameters and can be chosen to satisfy (16), with a constant independent of . (16)✦
For every there is a real constant such that the universal coordinate dimension satisfies (17) for every integer . (17)✦
In the paper, the result reads: with , the signed compound Poisson sum of the ramp jump belongs to , its th moments weighted by approach a target after averaging the nearly additive moment error, for constants the scale gives, for a scalar , , and, for each , discarding the whole vector whenever , where , gives a bounded gradient with -weighted error at most for sufficiently large and .
What Is Checked
The dimension estimate concerns the universal realization property in (2) for and , while the moment and cutoff estimates concern probability laws on finite real coordinate spaces and normalized gradients as in (4).