The Setting
Let be a finite set of indices and a standard Borel space. A family of random variables is -wise independent when, for each , the family is independent. The item concerns a probability measure on with marginals whose coordinates are -wise independent, for at least three indices; for this makes a pairwise independent coupling of , and .
The circle carries the Borel structure of ✦, and the circle group of is the group of measurable maps under pointwise multiplication.✦ An exact tuple for is a tuple in whose product is almost surely constant: for some , (1)✦ The exact tuples form the set .
The item is about how rigid exact tuples are, up to constant phases: how two of them compare at each coordinate, and how far one coordinate determines the rest.
Lines and Projective Tuples
For a measure on , the phase subgroup of consists of the maps that are -almost everywhere equal to a constant✦, and the projective circle group of is the quotient (2)✦ The line of is its class in . Two lines and in are orthogonal when , which does not depend on the choice of representatives.
The projectivization of tuples is the map (3)✦ and the exact projective tuples are its values on exact tuples, . The projection to coordinate takes a projective tuple to its th entry, and the exact lines at coordinate form its image .✦ The partners of an exact line at coordinate are the entries, at the other coordinates, of the exact projective tuple whose th entry is that line.
Codes
Exact lines are labelled by codes, which fix a vector of by its distances to a countable family of test functions. Fix a sequence of sets generating the -algebra of ; the test algebra is the algebra of sets it generates.✦ A test key is a finite list of triples with and ✦, and its test function is (4)✦ The test keys form a countable set; fix an enumeration of it and let be the test function of . For a finite measure on , the test vectors of are the classes of in , in this order.✦
A distance code for a matrix of reals is a sequence of nonnegative reals with (5)✦ Let be a kernel from a measurable space to , that is, a family of finite measures on depending measurably on . A code over is a pair of a point , the base point of , and a distance code for the distances of the test vectors of .✦ The vector of is a with for every .✦ The approximation index of is the least with ✦, and the representative of is the pointwise limit of its approximants , (6)✦ The distance of two codes and , with the base point of , is (7)✦
A vector of a Hilbert space is canonical for a sequence of vectors when the first not orthogonal to has a positive inner product with it: for some , (8)✦ A code over with base point is canonical when is canonical for the test vectors of ✦, and unimodular when holds -almost everywhere.✦
The Claim
Theorem 1 (Exact lines). Let be a finite set and a probability space.
Suppose . Let be -wise independent random variables on with values in measurable spaces , and let be measurable maps with almost surely, such that and are almost surely constant. Then, for some phases , (9)✦
Let be -wise independent complex random variables on with almost surely and almost surely constant. Then (10)✦
If and nonnegative reals satisfy for every , then (11)✦
Suppose now that and that is a probability measure on with marginals and -wise independent coordinates. If and for one , then (12)✦ The projections are injective: (13)✦ and for all , (14)✦ If and , then (15)✦ Finally, let and be tuples of canonical unimodular codes over such that and have the same base point and for every , and such that . Then (16)✦
In the paper, the result reads: if there are coordinates and any of them are independent, then distinct exact lines at a coordinate are orthogonal, an exact line has a unique tuple of partner lines, the exact tuples of lines form an abelian group whose projection to any coordinate is an isomorphism onto the exact lines there, these groups are countable when is separable, and nonconstant exact lines have mean zero.
The box gives the orthogonality as the first alternative of (9), the unique partners as (13), the isomorphisms as (14) and the zero mean as (15).
What Is Checked
The claims are stated over a finite index set , with at least three elements except in (10): over -wise independent random variables on a probability space in (9) and (10), over nonnegative reals in (11), and otherwise over a probability measure on with marginals and -wise independent coordinates, the circle group , the projective circle groups , the exact tuples , the exact projective tuples with their projections , and canonical unimodular codes over with their representatives and distances.
The library proves that the measurable maps form a subgroup of the group of all maps ✦, that and are subgroups of and of , and that and each are homomorphisms of groups, which makes a subgroup of .