The Setting
Let be the incidence graph of 13 points and 22 faces, each face a triple of points. Its vertex classes are the points and the faces , and a face has the points in its slots , , . The 66 edges of are the corners: a corner is a face with one of its slots , at the point .✦ Sidorenko’s conjecture asks of that for every symmetric kernel of mean one. The 22 faces, for in this order, are the triples , , , , , , , , , , , , , , , , , , , , , , each listed as .
The oriented kernels of this item are functions on a product of two finite probability spaces; they send the points of to and the faces to , and need not be symmetric. The item reaches them from a construction with random types: each point and each face of gets a type, and each corner a random activation, a set of active corners with a sign at every corner, whose law depends on the types at the two ends of the corner. Its ingredients are exponential priors on the types, a Gibbs thinning of these laws, dilution by an inactive type, and the crossed product, which turns a kernel on into a symmetric one.
Laws on Activations
A probability law on a finite set is a function with ✦; the expectation of a real function on under it is .✦
An activation of a finite set is a pair of a set of active elements and a vector of Boolean signs✦, and the Boolean monomial of a set is .✦ The activation character of a set is a real function on activations of , and the activation moment of a probability law on activations of at sets is a real number; they are given by (1) The inactive law is the point mass at the activation with no active element and every sign equal to .✦
For a nonempty and a bias , the tilted law on is the probability law on sign vectors that gives the weight .✦ The branch law of a set is the law on activations that is the product of the point mass at and this tilted law; its active set is always .✦ For sets and nonempty and a bias , the base activation law with bias is the probability law on activations of given by the even mixture (2)✦
Corners and Coefficients
The three sides of a face are the pairs , and of its points, in its slots , , ; the 33 point pairs are the pairs of points that are sides of faces.✦ A corner lies on the two sides of its face through : its left pair is the first of them✦ and its right pair the second.✦ The corners of a point pair are the corners with as left or right pair, the four corners at the two points of in the two faces containing .✦
For a corner , the corner union and the corner symmetric difference are the union and the symmetric difference of the set of corners of and the set of corners of , and the corner sign is at the corner , in slot of face , and at every other corner.✦ The corner law of is the base activation law on activations of the 66 corners.✦ The base pair law of a point and a face is the law on activations of the 66 corners given by (3)✦
A labelling gives each of the 33 point pairs a subset of a finite set ; it is nonempty when no pair gets the empty set.✦ Let give each corner a real function of two subsets of . The label product of at is the product over the corners of at the labels of and ✦, and the active coefficient of is the sum of its label products over the nonempty labellings.✦
Let the point types form a finite set and the face types a finite set , and let be a law on activations of for each point type and face type . The typed coefficient of type maps and is the active coefficient of the activation moments at the corners, (4)✦ the sum running over the nonempty labellings. For laws on and on , the type map law is the product of the product laws that give the points independent types drawn from and the faces independent types drawn from .✦ The averaged coefficient is the expectation of under the type map law.✦
Priors and Thinning
A score system on the type sets and consists of real scores for the point types , for the face types , and for the pairs of types.✦ The total score of type maps and is the real number (5)✦
For a real function on a finite set and a real number , the exponential prior is the probability law on proportional to , with normalizer .✦ For a score system on and and real numbers and , the Gibbs thinning factor of a point type and a face type is the number , as in (6): (6) The thinned law of a law on activations by a factor is the mixture of and the inactive law✦. For a score system on and , real numbers and and , the Gibbs prefactor of is the number .✦
Kernels and Their Moments
Let and be finite sets with probability laws and , and a real function on . The kernel mean of is , and the bipartite moment of is the homomorphism density of in with the points sent to and the faces to : (7) The transpose is a function on , and its bipartite moment draws the points from and the faces from . The crossed kernel of is the function on given by (8)✦
The Claim
Theorem 1. Let be a score system whose point types are the points and whose face types are the faces, and let and be real numbers with for every point and face . Take the exponential priors on the points and on the faces, and let be the base pair law thinned by . Then (9)✦
Theorem 2. Let and be real functions on a finite set , and let satisfy for every and . Then there is with (10)✦
Theorem 3. There are probability laws on the points and on the faces, taken as point and face types, and for every point and face a law on activations of the 66 corners with for every nonempty set of corners, whose averaged coefficient is negative: (11)✦
Definition 1. The diluted law of a probability law on a finite set , for , is the probability law on , with a new inactive type, given by (12)✦
Definition 2. The diluted pair law of a family of laws on activations of a finite set , indexed by and , is the family indexed by and given by (13)✦
Theorem 4. Let and be real numbers. There is with and (14)✦
Theorem 5. There are finite nonempty sets and , probability laws on and on , and an oriented kernel on them: a function with (15)✦
Theorem 6. Let and be probability laws on finite sets and , and a real function on . With the law on both sides, the crossed kernel of satisfies (16)✦
What Is Checked
The claims are stated over the incidence graph with its corners and their left and right pairs, probability laws on finite sets, the corner laws and base pair laws (3) on activations of the corners, the typed and averaged coefficients (4), the exponential priors and Gibbs thinning factors (6) of a score system, and the bipartite moment (7) and crossed kernel (8) of a function on . Of these constructions the library also proves that the corner symmetric difference is nonempty for every corner .✦