The Setting
A quadratic deficit compares the weight of a product construction with its weighted squared centered score. The Gaussian calculation uses continuous tent tests on the leaf coordinates and selector weights on a distinguished coordinate. The library fixes an explicit finite grid for this calculation.
Let be the standard Gaussian law on , and let have law . For a measure on and a real weight , the weighted mass is its integral in (1). (1)✦ The weighted mean is the normalized first moment in (2). (2)✦ The weighted variance is the normalized second moment minus the squared mean, as in (3). (3)✦
Weighted Tests and Selectors
The tent function with real center and width parameter is defined by (4). (4)✦ For a natural grid half-size , the grid index set is , with middle index . Clipping to is the real function in (5). (5)✦ For real spacing and , the leaf test center and the grid ramp difference are given by (6). (6) The selector partition weight adds the middle-index correction in (7). (7)✦
The fixed grid half-size and leaf tent width are the constants in (8). (8) The grid radius is fixed by (9). (9)✦ The grid spacing is the radius divided by the half-size, as in (10). (10)✦ The leaf indices form .✦ For , the leaf tent test is the function in (11). (11)✦ The selector test is the fixed-grid partition weight in (12). (12)✦
The Gaussian leaf mass is the weighted mass in (13). (13)✦ The Gaussian leaf mean is defined by (14). (14)✦ The leaf variance , total leaf variance , and remaining leaf variance are the quantities in (15). (15) The Gaussian deficit kernel is the weighted sum in (16); its Gaussian integral is in (17). (16)✦ (17)✦
Moment Gain and Finite Deficits
For an index set , a measure on a measurable space , a real score on and real weight functions on indexed by , the weighted moments of orders are the integrals in (18). (18)✦ The Gaussian leaf moment array and Gaussian selector moment array use the identity score, as in (19). (19) For real moment arrays indexed by , where is finite, the weighted means and variances are defined by (20). (20) The moment gain is the real quantity in (21), where all sums and products range over unless restricted. (21)✦
For a finite alphabet and a real function on , the uniform mean is the real number in (22), with value zero for the empty alphabet. (22)✦ For a finite alphabet , a finite leaf set , a real score on and real weights on indexed by , the finite weighted moment array is the array indexed by given in (23). (23)✦ The soft event weight associated with real selector weights and leaf weights is the real function on in (24), where . (24)✦ For a finite alphabet , a finite leaf set , a real score on , a real weight function on and a real centering value , the quadratic deficit is the real number in (25), using the uniform mean on . (25)✦ The leaf centering value is the sum of the weighted leaf means minus one, as in (26). (26)
The Claim
Theorem 1. For the fixed Gaussian grid and weights, the estimates (27), (29) and (30) hold, with (28) valid for every . (27)✦ (28)✦ (29)✦ (30)✦
In the paper, the result reads as (31) for leaf means , conditional variances , masses and selector , with parameters chosen so that and : (31)
What Is Checked
The Gaussian statement is over the standard Gaussian law, the fixed finite grid, and its tent leaf tests and selector tests. For every natural half-size and real spacing , the selector weights satisfy (32) at every real . (32)✦ The Gaussian moment arrays satisfy the gain identity (33). (33)✦
For a finite alphabet and finite leaf set , let have uniform mean zero and second moment one, and let be real weights with for every ; the soft deficit satisfies (34). (34)✦