The Setting
The centroid lengths and probabilities of the cells of a Gaussian partition describe its energy and the distribution of its mass. Normalizing the centroid lengths removes the scale of the energy, while one-dimensional Gaussian tails give scalar cell profiles.
Let be standard Gaussian probability measure, with mean zero and identity covariance, on Euclidean space .✦ The Gaussian centroid of a set is the unnormalized vector in (1); the centroid energy of a finite family is the sum of squared centroid lengths in that equation. (1)
A Gaussian measurable partition of is a family of Borel sets such that -almost every point belongs to exactly one cell.✦ Empty cells are allowed.
The active label set of a finite family consists of the labels of its positive-mass cells, and its active count is the cardinality in (2). (2) The probability of a cell and, for positive energy , its normalized active radius are defined by (3). (3)
An optimal Gaussian partition is a Gaussian measurable partition whose energy is at least that of every Gaussian measurable partition with the same labels in the same space.✦ A minimal optimal partition is an optimal partition whose active count is no larger than that of any other optimal partition with those labels in that space.✦
One-dimensional Tails
The unnormalized Gaussian density on and its upper tail are the functions in (4). Write for the Gaussian normalizing constant.✦ (4)
The Claim
Definition 1. For a finite family of subsets of with positive energy , the normalized active radii satisfy (5). (5)✦ For any Gaussian measurable partition , the active probabilities satisfy (6). (6)✦ For a minimal optimal partition with and positive energy, each active cell satisfies (7). (7)✦
The standard Gaussian density and distribution function on are defined by (8) and (9). (8)✦ (9)✦ The Gaussian quantile is the inverse of the standard Gaussian distribution function on , as in (10), and is zero outside that interval. (10) On , (11) and (12) define the Gaussian upper tail and inverse Mills ratio . (11)✦ (12)✦ The inverse Mills derivative in the upper-tail variable and its reflection in the variable are defined by (13). (13)✦ The Gaussian loss profile is the function defined by (14). (14)✦ For and with , its quantile expression is (15). (15)✦
In the paper, the result reads: for a hypothetical minimal optimal partition with active cells and value , where , in its -dimensional centroid span, the quantities and satisfy , and , with , and .
What Is Checked
The radius identity is stated for finite families of subsets of Euclidean space with positive centroid energy; the probability identities concern Gaussian measurable partitions, with minimal optimality, at least two labels and positive centroid energy for the upper bound. The quantile loss identity is stated for an arbitrary real tail threshold whose probability lies in .
The Gaussian distribution function in (9) is a strictly increasing bijection from onto .