The Setting
For a nonnegative integer , the discrete cube is .✦ A sign-valued function takes only the values and .✦ The question concerns the information retained by such a function after independent noise in each coordinate.
A real field is increasing if whenever for every coordinate .✦ An interior field takes its values in .✦ The uniform cube expectation of a real field is the normalized sum in (1). (1)✦
For real bias , the binary entropy deficit and binary entropy in natural logarithmic units are the functions in (2), where . (2) For a real field on , its mean binary entropy is the expectation of ✦, and its information deficit is the entropy Jensen gap defined in (3). (3)✦
The noise kernel at real correlation and the noise operator on real fields are given by (4). For , the coordinate flip probability is . (4) The notation in (4) parametrizes noise by time .
Dissipation and Its Scalar Profile
For and a coordinate , the reflected point is obtained by reversing and leaving every other coordinate unchanged.✦ For a real field and a coordinate , the discrete derivative is the field given by the half-difference across reflection in coordinate , and the entropy dissipation is the sum of expectations in (5). (5) The mean variance of a real field is the complement of its squared mean, as in (6). (6)✦
For , the inverse entropy deficit is the unique with .✦ The scalar dissipation rate and its dissipation excess are the functions of the deficit coordinate in (7). (7) The monotone correction is the function of a real entropy coordinate defined in (8). The small deficit threshold is specified in the same display. (8)✦
The Claim
Theorem 1. Fix the dimension .
Suppose every increasing interior field on satisfies (9), and every interior field on with satisfies (10): (9) (10) For every nonconstant increasing sign-valued function on and every , the conditional noise bound is (11): (11)✦
For every nonconstant increasing sign-valued function on and every , the noise bound (12) holds without the dissipation assumptions of the first assertion: (12)✦
Let be any sign-valued function on and let . There exists an increasing sign-valued function on with the mean and noisy information specified in (13): (13)✦
In the paper, the result reads: integrating gives , and continuity at zero time and reflection for negative correlation give Theorem 1.1.
What Is Checked
The conditional assertion quantifies its dissipation assumptions over interior fields on the finite uniform cube, the two noise bounds concern nonconstant increasing sign-valued inputs at , and rearrangement concerns arbitrary sign-valued inputs at .