The Setting
Take points and together with a finite family of vectors in a real inner-product space with norm , and fix scale bounds . To approximate a constant multiple of the Euclidean distance by a kernel integrated over scales in , one combines a Gaussian decay term in with trigonometric perturbations along the vector family. The question is which positive constant multiplies , and what explicit error bounds hold for truncating the underlying profile integral and for the base kernel distance on .
Definitions
The profile function is the real-valued function mapping each to (1)✦
On pairs , the scaled profile is the two-variable function defined by (2)✦
From the profile function , the profile constant is the real number given by (3)✦
For a real inner-product space with inner product and norm , a finite index set , a vector family , real weights , a scale , points , and coefficient functions , the kernel pairing is the real scalar defined by (4)✦
The Hilbert kernel for a vector family in a real inner-product space indexed by a finite set , real weights , integration bounds , points , and coefficient functions is the real number defined by (5)✦
The Claim
Theorem 1. The profile constant is strictly positive: (6)✦ For all real numbers satisfying and , (7)✦ Given a real inner-product space , a finite index set , a vector family , real parameters and , and points , letting denote the identically zero coefficient function on , (8)✦
In the paper, the result reads: for a large positive integer and a map on into a normed space, the constant (written in the paper) satisfies and whenever satisfy . The library answers the paper’s bound through (6) and (7) together with (8), using the kernel difference in place of while keeping , , , , , and unspecialized.
What Is Checked
Claims (6), (7), and (8) are stated over the profile constant defined from the profile function , together with the scaled profile at parameters and and the Hilbert kernel built from the kernel pairing for a finite vector family in a real inner-product space , weights , integration bounds , points , and zero coefficient functions.