The Setting
Take a standard Borel space with a measurable automorphism , a finite index set with elements, and an array object over initial array laws of , that is, a system of laws of arrays indexed by on a space of states, with a space of parameters. Over a parameter , a line is a circle-valued function on the states with parameter , up to a constant phase; over an array of level , a point of , a tuple of lines, one at each entry, is exact when the product of its functions is almost surely constant. The question is how the laws of the parameter arrays of lift to measures on labels of lines, on exact tuples of labels and on square arrays of labels whose rows and columns are exact, and how a label moves from one entry of an exact tuple, that is, one index in , to another.
An array of level over a space is a point of , an entry for each index tuple of length in ✦, and an array of level is a point of ✦; an axis of an array of level is a position in its index tuples, a slot is an element of , and the plane restriction fixing axis at slot takes an array of level to the array of level of its entries with inserted at position .✦ A family of probability measures on is consistent if every plane restriction carries to , and proper if, for every proper subset of , the last-axis planes at the slots in , the plane restrictions fixing the last axis at those slots, are independent under , each with law . Initial array laws of are consistent and proper laws on , invariant under applied entrywise.✦
An array object over consists of standard Borel spaces of states and of parameters, each with an automorphism, a factor map from to and a parameter map intertwining them, and invariant, consistent and proper laws on carried to by the factor map applied entrywise; it is moreover conditionally local, the last-axis planes at the slots of a proper subset of staying independent given the parameter array, each with the conditional law of an array of level given its parameters, and locally independent, the factor and parameter maps being independent under .✦
Applying entrywise carries to the parameter law on ✦, and the own kernel is the conditional law under of an array of level given its parameter array.✦ An array of level is the tuple of its entries ✦, its parameter array the tuple of their parameters ✦, and the raw tuple kernel of is the kernel from to that sends a parameter array to , the conditional law of the tuple of entries given . For a slot , the coordinate kernel sends to the conditional law under of given .✦
For every level , an array object carries data fields on : the parameter field and, for each index in , the outside field and the enlarged field , each a sub--algebra of the Borel sets. The projection onto a data field is the conditional expectation onto it, the orthogonal projection of onto the functions measurable for that field.✦
An extension of by an array object over the same initial laws is a pair of measurable maps, from the states of to those of and from the parameters of to those of , intertwining the automorphisms and compatible with the factor and parameter maps, the state map applied entrywise carrying each law of to the law of of the same level.✦ The array object is saturated if, for every extension of by an array object , every level and every data field , or on , the pullback isometry from to of the law of , with the state map applied entrywise, intertwines the orthogonal projections onto the functions measurable for that field.✦
Labels and Exact Tuples
Labels are built from codes of vectors. For a finite measure on , the Hilbert test sequence of runs, in a fixed enumeration, through the simple functions with rational and sets in the algebra generated by a fixed countable generating family of the Borel sets of , taken as vectors of .✦ A Hilbert code for is a pair of a base point and a distance code for the test sequence of : a nonnegative real sequence with and for all and , and with a term below for every .✦ A Hilbert code for decodes to a vector of at distance from for every ✦, and the representative of is the pointwise limit, as , of the simple function at the first index with ✦. The encoding of a vector of is the code .✦
A Hilbert code is canonical if the first test vector not orthogonal to its decoded vector has a positive real inner product with it✦, and unimodular if almost everywhere for .✦ Write for the space of Hilbert codes for . A label is a pair of a parameter and a code , a point of .✦ A label tuple is a pair of a parameter array of level and a tuple of codes, with label at slot ; label tuples thus form .✦ A label square is a pair of a parameter array of level and a matrix of codes, with labels , and its row and column are the label tuples obtained by fixing its first axis at or its last axis at ; label squares form .✦
Fix a kernel from to . The exact code set is the set of label tuples whose codes are canonical, unimodular and based at , and whose exact code mean has modulus one: (1)✦ The members of the exact code set are called exact code tuples. The line graph at a slot is the set of pairs of a parameter array and a code that is the entry of some ✦, and the exact square set is the set of label squares whose rows and columns all lie in .
For a slot , a matching assigns to each parameter array and each code for which one exists an exact code tuple over with entry at slot .✦ The label transition kernel on takes a label from slot to slot : at a label it is the law of the label at slot of the matched tuple over a parameter array drawn from , (2)✦
Counting Measures and Joining Operators
A Borel fibre enumeration of a set , for standard Borel spaces and , is a sequence of measurable sets and measurable maps such that for , every point of is for some with , and for and .✦ The counting measure of a Borel fibre enumeration over a measure on is the measure that is the image, under , of the product of with counting measure on restricted to : (3)✦
A set of labels matches the line graphs if, for -almost every and every slot , the codes with are those with . For such a set and Borel fibre enumerations of , of and of , the counting measures (3) over the parameter laws are on labels, on label tuples and on label squares.
For measure-preserving maps from to and from to , the pullback is an isometry of into ✦, and the joining operator of and is the bounded operator (4)✦ which sends to the conditional expectation of given , read on . When the slot maps carry to , is the joining operator (4) on of the slot maps at and at .
The Claim
Theorem 1 (Counting lift). Let be an array object whose index set has at least three elements, and let be a kernel from to such that, for every and every slot , the coordinate has law under and the coordinates with are independent.
Suppose is saturated and is the raw tuple kernel at for -almost every . Then there are a measurable set of labels , each with canonical, unimodular and based at , and Borel fibre enumerations of , of and of , such that matches the line graphs and, for all , (5)✦
Let be a set of labels matching the line graphs, and let and be Borel fibre enumerations of and of . For every slot , the parameter array of a label tuple has, under , the conditional law given its label at (6)✦
If moreover is saturated, is the raw tuple kernel at for -almost every and is a Borel fibre enumeration of , then for all slots and and all slots , the pair of labels of a label square has, under , the conditional law given the row (7)✦ with the transition kernel (2), and the joining operators of the slots satisfy (8)✦
Let , and be -finite measure spaces, with maps carrying to and carrying to for , such that for all ; let be a Markov kernel on and . If, under , the pair has conditional law given , then (9)✦
In the paper, the result reads: there are -finite measures on and on counting matched label tuples and arrays, with marginals and , and transition kernels whose operators are contractions with adjoint commuting with time, satisfying for , and proper families of labelled columns have a product conditional law given a labelled row.
What Is Checked
The claim is stated over array objects over initial array laws, with their extensions and saturation; the parameter laws , the own kernels and , the raw tuple kernel and the coordinate kernels ; Hilbert codes for , their representatives and the canonical and unimodular codes; the exact code set of (1), its line graphs and the exact square set ; Borel fibre enumerations and their counting measures (3); the transition kernels of (2); and the joining operators (4).
About these constructions the library proves that the slot maps on label tuples are measurable✦ and that, under the hypotheses of (6), for every set of labels matching the line graphs and all Borel fibre enumerations of it and of , each slot map carries to .✦