Erdos-Szekeres-type statements: Ramsey function and decidability in dimension 1
Abstract
A classical and widely used lemma of Erdos and Szekeres asserts that for every n there exists N such that every N-term sequence a of real numbers contains an n-term increasing subsequence or an n-term nondecreasing subsequence; quantitatively, the smallest N with this property equals (n-1)2+1. In the setting of the present paper, we express this lemma by saying that the set of predicates Phi=x1<x2,x1 x2$ is Erdos-Szekeres with Ramsey function ESPhi(n)=(n-1)2+1. In general, we consider an arbitrary finite set Phi=Phi1,...,Phim of semialgebraic predicates, meaning that each Phij=Phij(x1,...,xk) is a Boolean combination of polynomial equations and inequalities in some number k of real variables. We define Phi to be Erdos-Szekeres if for every n there exists N such that each N-term sequence a of real numbers has an n-term subsequence b such that at least one of the Phij holds everywhere on b, which means that Phij(bi1,...,bik) holds for every choice of indices i1,i2,...,ik, 1<=i1<i2<... <ik<= n. We write ESPhi(n) for the smallest N with the above property. We prove two main results. First, the Ramsey functions in this setting are at most doubly exponential (and sometimes they are indeed doubly exponential): for every Phi that is Erdos--Szekeres, there is a constant C such that ESPhi(n) < exp(exp(Cn)). Second, there is an algorithm that, given Phi, decides whether it is Erdos-Szekeres; thus, one-dimensional Erdos-Szekeres-style theorems can in principle be proved automatically.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.