The Setting
Take a stationary random array of states over a finite index set , each state carrying a parameter, and cut a square from it: rows and columns indexed by , each row and each column with its conditional law given its parameters. A circle-valued function of a block of cells may factor almost surely along the rows of the block, into functions of the other columns, or along its columns, into functions of the other rows. The question is when a function of a single cell that factors along its row also factors along its column, and how factoring along rows sits with the Markov operators that pass between rows along the columns.
Let be a measurable automorphism of a standard Borel space , a nonempty measurable space isomorphic to a Polish space with its Borel sets.✦ An array of order over a space is a map from to , a point of ✦; an array of order zero is a point of ✦. For an array of order , an axis and an index , the plane of fixing axis at is the array of order whose entry at a tuple is the entry of at the tuple obtained by inserting into at position .✦ Initial array laws of are probability measures on , one for each , that are invariant under applied entrywise, consistent, in that every plane of an array drawn from has law , and proper: for every proper subset of , the planes fixing the last axis at the indices in are independent under , each with law .✦
An array object over initial array laws of consists of a standard Borel space of states and a standard Borel space of parameters, each with a measurable automorphism, measurable equivariant maps and , and probability measures on that are invariant, carried to by applied entrywise, consistent and proper in the same sense, and conditionally local: given the parameter array, obtained by applying entrywise, the planes of properness stay independent, each with the conditional law of an array of order given the corresponding plane of the parameter array; at order zero, and are independent under .✦
The parameter law of is the law of the parameter array of an array drawn from , a probability measure on .✦ The kernel is the Markov kernel from to that gives at the conditional law under of the array given that its parameter array is .✦
An extension of by an array object over the same initial laws is a pair of measurable equivariant maps, from the states of to those of and from the parameters of to those of , with and , such that applied entrywise carries to for every ✦; pullback along it sends in the complex space to in , with applied entrywise✦.
The data fields of on are the parameter field , the sub--algebra of the Borel sets generated by the parameter array, and, for each , the outside field and the enlarged field at the index , sub--algebras of the Borel sets of ; for a data field , is the conditional expectation onto , the orthogonal projection of onto the functions measurable for . The array object is saturated if, for every extension of by an array object , every and every data field , (1)✦
The Square
An array of order is read as a square. Its row is the plane fixing the first axis at , its column the plane fixing the last axis at , both of order , and its cell is the array of order obtained by fixing both. For a set of rows and a set of columns, is the block of cells in those rows and columns✦; and are the cells of column in the rows of and of row in the columns of . A parameter array of order is read in the same way. At , and , points of , are the parameters of row and of column .
The law of cells in the rows of and the columns of at a parameter array is the law of the block when the rows , , are drawn independently, row from .✦
A complement split of a nonempty set is an ordered pair of nonempty subsets of such that , and are pairwise disjoint with union .✦ A triple partition of is an ordered triple of nonempty, pairwise disjoint subsets of with union .✦
Let , , be sets of rows, a set of columns and a family of cells in the rows of and the columns of . Draw each column , , independently from conditioned on ✦; this couples the blocks and ✦. The column operator is the joining operator of this coupling: the bounded operator from to that sends a function of to its conditional expectation given .
The column-active space is the essential range of the family with respect to , the orthogonal complement in of the essential kernel of the adjoint family: (2)✦
A measurable map to the unit circle of , with its Borel sets, defines a vector of of a finite measure, written again.✦ Let be a complement split of the set of columns. A vector of is exact for the row split if with and a measurable map from the blocks of cells in the rows of and the columns of to such that, for some measurable maps and to , (3)✦ The set is the set of vectors of that are exact for the row split .✦
A vector of is fully exact, exact for the law of the rows of and for the law of the columns of , if with and a measurable map from the blocks of cells in the rows of and the columns of to such that, for some constants and measurable maps , , and , , to , (4)✦ The set is the set of fully exact vectors of .✦
Cells and Lines
At , a block with the single row and the single column is one point of , and is the measurable isomorphism from such blocks onto .✦ For a measure on to which carries , a vector of pulls back to in .
A family of measurable maps from to is an exact tuple for a law on if, for some constant , (5) For a finite measure on , a slot and , the exact cone is the set of vectors of with and an exact tuple for .✦
For a measure on , a slot and , the slot-active space is the following closed subspace of .✦ Draw an array of order one from , with entries , . If the entry has law , then is the closure of the range of the joining operator from of the law of the other entries to , the operator that sends a function of to its conditional expectation given ; otherwise .
A line is a measurable map from to up to a constant phase, and lines are labelled by codes. The test vectors are the complex rational combinations of indicators of sets in the algebra generated by a fixed countable generating family of the -algebra of , taken in a fixed enumeration.✦ A Hilbert code for is a pair of a base point and a sequence of nonnegative reals that locates one vector of by its distances to the test vectors.✦ The representative of a Hilbert code for is the pointwise limit, as , of the first test vector within of the vector that locates.✦ A Hilbert code for is unimodular if almost everywhere for ✦, and canonical if the first test vector not orthogonal to the vector that locates has a positive real inner product with it.✦
For a Markov kernel from to and a slot , the line graph is the set of pairs of and a code for such that for an exact family of codes at : canonical unimodular codes , , each with base point , with (6)✦
The Claim
Theorem 1. Let be an array object over initial array laws of , with finite index set , and let .
Let be saturated, let , let be a set of rows with a complement split , and let be a set of columns that has a complement split. Then for -almost every parameter array of order , (7)✦ the intersection running over all complement splits of .
Let be saturated, let have at least two elements, and let be a complement split of . Then for -almost every the following holds: if is a measure on to which carries and , then (8)✦
Let be a parameter array of order two such that the column of a square drawn from has law , and let be a finite measure on to which carries . Then for every , (9)✦
Let be saturated, let , and let and be triple partitions of , read as sets of rows and of columns. Then for -almost every the following holds: if , , are measurable maps to of the blocks of cells in the rows of and the columns of , , respectively with for -almost every , then (10)✦
Let be saturated, let have at least three elements, and let differ from . Then for -almost every and every exact tuple for , (11)✦
Let be saturated, let have at least three elements, and let , for , be maps from to with measurable for each , forming an exact tuple for for -almost every . Then for -almost every , (12)✦
Let be saturated, let have at least three elements, and let be a Markov kernel from to with for -almost every , under which, for every and every , the coordinate has law and the coordinates other than are independent. Then for -almost every and every Hilbert code for , (13)✦
In the paper, the result reads: for almost every parameter square and at every cell , the projections onto the paper’s subspaces and commute and their common range is spanned by lines exact for both the full row law and the full column law , the row Markov operators intertwine the column-active projections and the column operators the row-active ones, and for an exact row tuple and partner indices and in the row operator carries to an explicit multiple of , so that column activity of one partner line is equivalent to that of any other, and symmetrically.
What Is Checked
The claim is stated over a saturated array object with its parameter laws and conditional laws , the laws of cells of a square, the column operators and the column-active spaces , the sets and of vectors exact for a row split and fully exact, and, at a single cell, the slot-active spaces , the exact cones and the line graphs of canonical unimodular Hilbert codes.
Beyond the claim, the library proves that the vectors a family of operators sends to zero almost everywhere form a linear subspace✦, and that a measurable map to is square integrable for every finite measure✦. It also proves that pullback along an extension is a linear isometry at every order.