The Setting
An amplification block repeatedly assembles a binary word from a pair of binary words called atoms. At nesting level , let be the repetition count; the comparison is between deletion traces of and copies of the same atom, whose source lengths differ.
Let be the requested deletion probability, so each bit survives with probability . Put and . The construction theorems call their working deletion rate ; their rate is written here. A binary word of length assigns a bit to each position .✦ A deletion mask of a binary word is a choice of retained positions; the retained subsequence lists their bits in order. A complete binary trace is a finite bit string, including its surviving length.✦ The complete trace space consists of bounded traces with bound .✦
A finite probability law on a finite set is a nonnegative mass function with total mass one.✦ For a real test function on and a natural half order , the expectation and squared moment norm are defined by (1); the latter uses probability weights. (1) The point mass at is the probability law on that assigns mass one to .✦ For probability laws on a common finite outcome set, indexed by , the mixture of and is the probability law assigning mass to each outcome .✦ The pushforward of a finite probability law under a function is the mixture of and the point masses at .✦
The deletion mask law at rate assigns the weight in (2) to a mask.✦ The complete trace law of a binary word at deletion rate is the pushforward of the deletion mask law under , on any with .✦ The complete trace expectation of a test is .✦ (2)✦
The overlap of laws on a common finite set is the quantity in (3).✦ Complete trace overlap uses the common cutoff for source words .✦ (3)
An approximate trace simulation of a binary word with error consists of a finite law on and a decoder satisfying (4) for every trace test .✦ (4)
The rank Schur constant , primitive period constant , initial overlap lower bound , and overlap constant are the constants in (5). (5) For maximum block dimension and total block count , the warmup block count , block dimensions , block gain , total score exponents , and scale allowance are given by (6). Block index starts at zero and corresponds to block in the paper. (6) For a scale parameter , the scale exponent , window exponent , scale exponent product , and backwards target scale are the quantities in (7). Subtraction in a natural index is truncated at zero. (7) The half moment order and source error allowance are defined by (8); the even score exponent is . The notation means the natural ceiling truncated at zero. (8)
A concrete level pair at scale , level , error allowance , score decay exponent , overlap threshold , count multiplier , and half order consists of distinct words of a common length , two finite laws on a common outcome set, a common decoder, a real likelihood score , an error , and an integer repetition count . The laws simulate , respectively, in the sense of (4), and the remaining conditions are (9). The repeated word is the concatenation of copies of .✦ (9) A source level pair at level is a concrete level pair with parameters (10); write for its error. (10) with from (8) and from (5).
Words and Counts Inside a Block
Fix a target scale and block dimension . The integer window size and extended block windows are the sequences in (11), indexed by zero-based nesting level . Natural floors are truncated at zero, as indicated by the subscript . (11) A block length and overlap state is a quadruple of natural numbers recording word length, overlap count, copies, and trials; the last two describe the transition into that state.✦ Starting with the length and overlap count of , the nested arithmetic states satisfy (12).✦ Here are the first two entries of , is the root trial count at , and is the local trial count at copy count and overlap count . (12) and are the copy count and block trial count for the transition from level to .
For cancellation order , put and let be the Prouhet binary pattern. The recursive source block words are pairs of words defined by (13), where denotes complementary-word substitution. Its slot counts and nesting levels start at zero. (13)✦
Window Budgets and Numerical Margins
For window counts and depth , the complete window leaf set consists of all paths in the deterministic window tree, including unoccupied paths✦; its cardinality is the source leaf count.✦ The window collection set indexes the local and global trial collections at every nonleaf vertex.✦ Their recursions are (14), where and the two root tags denote local and global collections. (14) For local trial counts , the scheduled global trial counts obey (15). The collection trial count at depth is at the local root tag and at the global root tag; a collection below a child inherits its depth- count.✦ (15) The hidden metadata tail bound , fitting tail bound , and hidden binomial cap are the quantities in (16). (16) For cancellation order , the sparsity count , centered moment budget , and clipping cost are defined in (17). Their arguments include a half order , fitting threshold , and leaf-score norm allowance . (17)
A finite ordered role shape of depth and branching bound is an ordered rooted tree with at most children at each vertex and one of three roles attached to each child; at depth zero there is one empty shape.✦ The shape and role count is the number of these shapes, without numerical child labels.✦ The binary tree budget for level costs and the interior label budget are given by (18). (18) The forced contribution budget is the quantity in (19); all truncated exponents there are natural integers. (19)✦ For smoothing allowance , fitting threshold , cutoff , and score weight , the sharp degree budget and sharp energy budget are defined by (20). (20)
The source degree cutoff , work size , cancellation order , moment envelope , weight envelope , cap envelope , tail denominator , and tail exponent are the constants in (21). (21) The bad role factor , forced role factor , smoothing prefactor , and skeleton prefactor are specified in (22), with for the source cancellation order in (21). (22) The numerical schedule margins at target scale are the inequalities (23); the last two apply to every integer . (23)✦
The Claim
Theorem 1. Let , , and , and let be a source level pair at level and working rate . Put and . Assume the numerical schedule margins (23) at and the tail bound . Form the words and counts from by (12) and (13), using the source cancellation order. The adjacent-power estimate is (24). (24)✦ If , use the dimension-two block schedule and put , , and . Let and for every , and set . The second warmup estimate is (25). (25)✦
In the paper, the result reads: under the input hypotheses of Proposition 5.1 and the specified word and count construction, every satisfies , where is the positive nested atom.
What Is Checked
The claims are stated over finite binary words, finite approximate trace simulations with a probability-weighted likelihood-score moment bound, and the specified integer word and count schedules under the numerical margins and exponential tail bound. The deletion mask weights are nonnegative for rates in ✦, and their sum is one for every real deletion parameter.✦