The Setting
The question is whether a deficit in the small-scale entropy of a self-similar measure can coexist with a larger entropy rate of affine compositions. A weighted affine iterated function system on a finite alphabet consists of similarities with real signed slopes and translations, and positive normalized weights, as in (1). (1)✦ For a measure on , its weighted component is the weighted pushforward in (2); self-similarity means the equality there between and the sum of these components. (2)
A finite word has word probability and complete affine map given by (3); the empty word has probability one and the identity map. (3) The set of attainable maps at length consists of the distinct pairs for words of length .✦ The random walk mass of an attainable map is the sum of the probabilities of words with the same signed slope and translation as .✦ The random walk entropy , its entropy rate , and the positive Lyapunov exponent use base-two logarithms in (4). (4) Every finite word has strictly positive word probability .✦
The symbol depth is the base-two logarithm of the reciprocal contraction magnitude.✦ The contraction depths and of finite words are the real numbers given in (5). (5) For a function on finite words, its Bernoulli word average at length is in (6). (6)✦
Coding Information
The symbol probability measure on the finite alphabet , with all subsets measurable, is the weighted sum of point masses .✦ The Bernoulli address measure on is the countable product of the symbol probability measure .✦ For an address , the prefix contraction product scales the translation at position to the term ; the coding map and the self-similar coding measure are defined in (7). (7) The scaled translation term of an address is the prefix contraction product multiplied by the translation .✦
For measures on , their maximal ball mass ratio is the supremum in (8), over positive rational radii at most one; denotes the closed ball. (8)✦ For measures on , the positive logarithmic maximal envelope is the maximum of zero and the natural logarithm of the real value of the maximal ball mass ratio .
The symbol information observable is the negative natural logarithm of the initial symbol’s weight.✦ The logarithmic density correction is the natural logarithm of the real Radon-Nikodym derivative of with respect to its weighted component at .✦ The symbolic maximal envelope , the clipped correction , the projection information rate , and the coding dimension exponent are defined by (9). (9)
Entropy and Pair Capacity
For a measure and an observable , the fiber mass is the real value of .✦ The partition entropy of a measure and an observable is the sum in (10), with the contribution at zero mass defined to be zero. The shifted grid cell index gives the shifted grid entropy at scale and real offset , whose Lebesgue average over is . The dyadic deficit is defined in the same display. (10) These entropies use natural logarithms; division by gives base-two units.
A measure is compactly supported when for some .✦
A finite probability law on consists of a finite support with weights summing to one.✦ For an integer , its pairs at scale are the unordered pairs in in (11).✦ For a finite probability law on and an integer , a nonnegative pair weighting on is feasible when its incident pair mass is at most at every point of .✦ The fractional matching capacity of a finite probability law on is the supremum of the total feasible pair weight at scale . (11)
For a finite source set with weights and an observable , the fiber mass at is the sum of over .✦ For a real observable on a finite source set with positive normalized weights , the distribution of is the finite probability law assigning to each distinct value of the sum of over the source points satisfying .✦ For a real observable , positive source weights, and a tag map , the tag fiber , its mass , and the conditional observable distribution are defined by (12). (12) For strictly positive source weights, the conditional weights sum to one over the tag fiber .✦
Fix a map to a finite type alphabet. For a map and a word , the word tag sequence is the sequence in with entries .✦ The type-conditioned translation law is the conditional observable distribution of on , with weights and tag map , at the word .✦ For a finite list of symbols in , the type-conditioned translation law is the conditional observable distribution of on , with weights and tag map , at the word formed from in address order.✦
Retained Targets
For a real depth bound , an integer band interval , and a nonnegative block length , the averaged band capacity budget is the weighted sum in (13). (13)✦ All integer sums range over inclusive integer intervals and are zero when the interval is empty.
For a depth profile , an index is retained at target when the two conditions in (14) hold. (14)✦ For a finite word , write for its first letters and for the next letters, truncating either segment at the end of . The prefix depth profile of a finite word is the function . For a finite word , the retained matching capacity is the nonnegative real number in (15), determined by the prefix depth profile . (15)✦
The Claim
Theorem 1. Let and let be integers with , , and . Then (16)✦
Fix nonnegative integers with , a real , and integers with . Suppose that every finite word satisfies and . Then $$\sum_{L=1}^{T}\mathbb E_H\!left[C^{A',n,a,B}_{L,j}\right] =\mathcal B_{A'}(a,B;n).$$(17)✦
Suppose that symbols with the same type have the same signed slope, and let be a compactly supported self-similar probability measure for . Let satisfy for every . Let be nonnegative integers with , , , and for every symbol . Choose nonnegative integers and integer endpoints for , and a real . Assume that , , and for every , and that whenever . Then (18)✦
Suppose that symbols with the same type have the same signed slope, and let be a self-similar probability measure for . Let be real numbers with , , and . Suppose that for every . Then (19)✦
For every self-similar probability measure for , the coding dimension exponent satisfies (20)✦
In the paper, the result reads: summing targets gives a fixed positive contribution from each band, whereas the disjoint-window budget is , yielding with positive constants and block-depth bound , which is impossible for arbitrarily large , so and cannot hold together.
What Is Checked
The identities (16) and (17) concern integer target intervals and finite Bernoulli words; the bounds (18) through (20) concern self-similar probability measures and their finite type-conditioned translation laws, with compact support and a uniform deficit required in (18). Every finite word has strictly positive probability✦, and the conditional weights on each tag fiber sum to one.✦