The Setting
A finite observation of a sign cube retains part of the coordinate sum through its conditional score. The relevant quantities are the second moment of that score and the Fourier degrees of the observation’s fiber indicators. The resulting Boolean functions are measured by their signed singleton Fourier sums relative to the square root of their degrees.
For a finite coordinate set , the sign cube is the set of maps from to .✦ A Boolean function on a set is a function taking values in ; a function is nonconstant when it takes different values at two points. All finite averages are uniform. For a real observable on a finite set , its uniform mean is (1)✦
For , the Fourier character and the real Fourier-Walsh coefficient of are given by (2) For a real function on a finite sign cube, the ordinary real multilinear Fourier degree is the largest cardinality of a set with nonzero coefficient in (2), with degree zero for constants and the zero function. For a real function on a finite sign cube, the signed singleton sum is the real number given by (3), with the signs of the coefficients retained rather than replaced by their absolute values: (3)✦ The degree used with (3) is distinct from polynomial threshold degree and degree over a finite field.
An observation is a map into a report set . For an observation , a report and a real observable on the sign cube, the fiber indicator is the real-valued function equal to one on the fiber over and zero elsewhere, while the fiber probability and restricted moment are the real numbers given by the uniform averages in (4): (4) For an observation and a real observable on its sign cube, the fiber conditional mean is the function given by the quotient of the restricted moment by the fiber probability in (4), with value zero on an empty fiber. The conditional expectation of a real observable under an observation is the real-valued function on the sign cube obtained by evaluating the fiber conditional mean at the observed report: (5)
For an observation , the coordinate sum and the centered conditional score are real-valued functions on the sign cube, and the retained variance is the second moment of , given using (5) by (6) For an observation and a nonnegative integer , the grouped score scale is the nonnegative real number defined from the retained variance in (6) by (7)
Definition 1. The sign threshold is the function (8)✦
The score sign readout of an observation is the function from the sign cube to obtained by applying the threshold in (8) to the conditional score : (9)✦
The Gaussian Gain Constant
For real parameters and , the tent function is given by (10). (10)✦ Let denote the standard Gaussian probability measure on . The Gaussian gain test is the function and the gain constant is the number defined from (10) by (11)
For a real observable on a finite set and a nonnegative integer , the normalized sum is a real-valued function on , equal to zero when and given by (12) when . (12)✦
The Claim
Theorem 1. Let be finite and let have . Let be a nonnegative integer such that every fiber indicator has degree at most . With the scale in (7) and the constant in (11), there exist a positive integer and a Boolean function satisfying (13)✦ Suppose further that and that satisfies . There exist a positive integer and a nonconstant Boolean function satisfying (14)✦
In the paper, the result reads: if finite observations have positive cell degree bounds and arbitrarily large ratios , then Theorem 1.1 holds. The first assertion gives amplified Fourier gain: a Boolean function on a positive number of blocks has degree at most the number of blocks times the fiber degree bound, and its signed singleton sum exceeds the Gaussian gain constant times the grouped score scale.
What Is Checked
The assertions (13) and (14) are stated over uniform finite sign cubes and the ordinary real Fourier degrees of the fiber indicators of , with positive retained variance; (14) also assumes positive and and the stated ratio inequality.
For the readout in (9), the library states the Boolean-valuedness assertion (15) and the signed singleton identity (16): (15)✦ (16)✦ When every fiber indicator of has degree at most a nonnegative integer , the readout satisfies (17): (17)✦ The score sign readout has Fourier degree at most the common Fourier degree bound for the fiber indicators.
The Gaussian constant defined in (11) satisfies (18): (18)✦ For a real observable on a finite set with and , the normalized sums in (12) satisfy (19): (19)✦ The Gaussian gain constant is strictly positive. Every real observable on a finite set with uniform mean zero and second moment one has a normalized sum over a positive number of samples whose uniform mean absolute value exceeds the Gaussian gain constant.