Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Published pp. 259–267 (PDF). These definitions also fix conventions in the expanded proofs.

Ambient sizes in theorem statements are positive integers. A terminal empty ground set in a deletion proof is handled separately.

For an explicitly specified finite set XX, write Ω(X;k)={A⊆X:∣A∣=k}\Omega(X;k)=\{A\subseteq X:|A|=k\} and Ω(X;l)\Omega(X;\mathbf l) for ordered partitions of XX with cell sizes l=(l1,…,ls)\mathbf l=(l_1,\ldots,l_s). Its cardinality is ∣X∣!/∏li!|X|!/\prod l_i!. A family has no repeated members. Pairs and tuples in counting statements are ordered; no distinctness is imposed unless stated or forced by the pattern.

For 0<p<10<p<1, define

μp,X(F)=∑F∈Fp∣F∣(1−p)∣X∣−∣F∣,dX(F)=μ1/2,X(F)=∣F∣/2∣X∣.\mu_{p,X}(\mathcal F)=\sum_{F\in\mathcal F} p^{|F|}(1-p)^{|X|-|F|},\qquad d_X(\mathcal F)=\mu_{1/2,X}(\mathcal F)=|\mathcal F|/2^{|X|}.

The ground set is retained even if its points occur in no member. This is necessary for the slice identities used in the proofs. The source's notation p(F)p(\mathcal F) is written using ∣⋃F∣|\bigcup\mathcal F| on p. 267; we use the explicit current ambient ground set throughout the deletion argument, rather than silently removing unused coordinates.

For x∈Xx\in X, put F0={F∈F:x∉F}\mathcal F_0=\{F\in\mathcal F:x\notin F\} and F1={F∖{x}:x∈F∈F}\mathcal F_1=\{F\setminus\{x\}:x\in F\in\mathcal F\}, both on X∖{x}X\setminus\{x\}. Thus μp(F)=(1−p)μp(F0)+pμp(F1)\mu_p(\mathcal F)=(1-p)\mu_p(\mathcal F_0)+p\mu_p(\mathcal F_1). Write (F,G)∈P(X;[a,b])(\mathcal F,\mathcal G)\in\mathcal P(X;[a,b]) if every cross intersection avoids every integer in [a,b][a,b]. Directly from whether xx is present,

(F1,G1)∈P(X−x;[a−1,b−1]),(F0,G0∪G1)∈P(X−x;[a,b]),(F1,G0∩G1)∈P(X−x;[a−1,b]).\begin{aligned} (\mathcal F_1,\mathcal G_1)&\in\mathcal P(X-x;[a-1,b-1]),\\ (\mathcal F_0,\mathcal G_0\cup\mathcal G_1)&\in\mathcal P(X-x;[a,b]),\\ (\mathcal F_1,\mathcal G_0\cap\mathcal G_1)&\in\mathcal P(X-x;[a-1,b]). \end{aligned}

For partition families A(i)\mathcal A^{(i)}, the joint pattern is the array

mj1…jr=∣⋂i=1rAji(i)∣.m_{j_1\ldots j_r} =\left|\bigcap_{i=1}^r A^{(i)}_{j_i}\right|.

It is compatible with the specified cell sizes if its entries are nonnegative integers and its one-coordinate marginals are those sizes. Compatibility is equivalent to realizability: partition XX into labeled atoms of the given sizes, then take the indicated unions. Consequently, the number of full-family tuples realizing a compatible pattern is exactly

N(M)=n!∏j1,…,jrmj1…jr!.N(M)=\frac{n!}{\prod_{j_1,\ldots,j_r}m_{j_1\ldots j_r}!}.

The notation iMi_M counts such tuples in the chosen families. For two uniform set families, ili_l counts pairs with intersection size ll; the full-family count is

N(n;k,h,l)=(nk)(kl)(n−kh−l).N(n;k,h,l)=\binom nk\binom kl\binom{n-k}{h-l}.

The full proofs often use a density e−ϵne^{-\epsilon n} in place of (1−ϵ′)n(1-\epsilon')^n, where ϵ′=1−e−ϵ\epsilon'=1-e^{-\epsilon}. Statements using these two conventions are equivalent after renaming the positive constant. All auxiliary proportions rounded to integers are assigned explicitly; parameters written as cell sizes are always integers.