By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

All papers

Claims

Findings

A classification of finite Euclidean Ramsey configurations

A finite set A={a1,…,as}⊂RdA = \{a_1, \dots, a_s\} \subset \mathbb{R}^d of at least two points with affine span Rd\mathbb{R}^d is Euclidean Ramsey if and only if there is a matrix P∈Matd+1(B)P \in \mathrm{Mat}_{d+1}(B) such that (pi⊗1)TP(1⊗pi)=0(p_i \otimes 1)^T P (1 \otimes p_i) = 0 for all 1≤i≤s1 \le i \le s and mF(Pαβ)=δαβm_F(P_{\alpha\beta}) = \delta_{\alpha\beta} for all 1≤α,β≤d1 \le \alpha, \beta \le d, where pi=(1,ai)Tp_i = (1, a_i)^T, F=Q((ai)α:1≤i≤s,1≤α≤d)F = \mathbb{Q}((a_i)_\alpha : 1 \le i \le s, 1 \le \alpha \le d), B=F⊗QFB = F \otimes_{\mathbb{Q}} F, and mF:B→Fm_F : B \to F is the multiplication homomorphism.

If the rational tensor condition (3) has no solution for AA, then there is an integer r0≥2r_0 \ge 2 such that every RD\mathbb{R}^D admits an r0r_0-coloring with no monochromatic congruent copy of AA.

If for every integer n≥1n \ge 1 there exist two paths in KsK^s with the same endpoint and weights λ>0\lambda > 0 and μ≥0\mu \ge 0 satisfying μ/λ<1/n\mu/\lambda < 1/n, then AA is Euclidean Ramsey.

Every nonempty subset of a finite transitive Euclidean set is Ramsey.

Every nonempty set of at most five distinct points on a circle is Ramsey; in particular, every cyclic quadrilateral is Ramsey.

For every transcendental a∈(−1,1)a \in (-1, 1), the cyclic quadrilateral Ka={(−1,0),(1,0),(a,1−a2),(a,−1−a2)}K_a = \{(-1, 0), (1, 0), (a, \sqrt{1-a^2}), (a, -\sqrt{1-a^2})\} is Ramsey and is not subtransitive, disproving the necessity direction of the Leader–Russell–Walters subtransitive characterization and their non-Ramseyness conjecture for these kites.

For algebraically independent real numbers t1,…,t9t_1, \dots, t_9 over Q\mathbb{Q}, the nine points ai=(1−ti21+ti2,2ti1+ti2)a_i = \left(\frac{1-t_i^2}{1+t_i^2}, \frac{2t_i}{1+t_i^2}\right) (1≤i≤91 \le i \le 9) form a spherical set that is not Ramsey.

The twelve-point subset X=R+Q0∪R−Q0∪Q0X = R_+ Q_0 \cup R_- Q_0 \cup Q_0 of the unit circle, where Q0={(±1,0),(0,±1)}Q_0 = \{(\pm 1, 0), (0, \pm 1)\} and R±R_{\pm} are rotations by ±2arctan⁡t\pm 2\arctan t for Liouville's constant t=∑m=1∞10−m!t = \sum_{m=1}^{\infty} 10^{-m!}, is not Ramsey.

Methods

Algebraic separation of the symmetric rational tensor evaluation map in Y=(R⊗QR)s⊕Symd(R)\mathcal{Y} = (\mathbb{R} \otimes_{\mathbb{Q}} \mathbb{R})^s \oplus \mathrm{Sym}_d(\mathbb{R}) to construct a finite modular coloring avoiding congruent copies of AA in every dimension.

Source excerpts

“In the quotient vector space Y/ im Φ, extend the nonzero class of y0 to a basis over Q. Sending that basis element to 1 and the others to 0 gives a Q-linear functional ℓ : Y → R such that ℓ ◦ Φ = 0, ℓ(y0) = 1.” · Show in paper

Formal logarithmic coordinates and m\mathfrak{m}-adic completion in the Laurent polynomial ring Q[X1±1,…,Xk±1]\mathbb{Q}[X_1^{\pm 1}, \dots, X_k^{\pm 1}] to convert rational tensor equalities into finitely supported lattice cancellation identities.

Source excerpts

“To use the tensor hypothesis, we pass to formal logarithmic coordinates, in which the generators Xh − 1 become linear forms multiplied by units. Put m = (X1 − 1, . . . , Xk − 1) and complete R with respect to m.” · Show in paper

Gaussian covariance-interpolation smoothing, discretization, and rational perturbation within a fixed convolution image to produce two-scale configurations with identical coordinate multisets.

Source excerpts

“We will average Gaussian densities with covariance matrices running from S to S + C.” · Show in paper

Ultrafilter semigroup corner group p(βS)pp(\beta\mathcal{S})p and simultaneous realization of ultrafilter products by finite strings, combined with the Hales–Jewett theorem, to convert equal-endpoint free-group paths with arbitrarily small weight ratios into monochromatic congruent copies.

Source excerpts

“We instead map K into a group of ultrafilters on S and then realize its common factors by common strings.” · Show in paper

Permutation averaging over finite coordinate-permutation orbits in the free group KK to equalize path endpoints at a weight cost small relative to the path degree.

Source excerpts

“Process the finitely many orbits, increasing the derived depth on each to s − 2. Every degree stays positive, and we obtain membership in L for all within-orbit differences.” · Show in paper

Derivation-based weighted moment obstruction on the coordinate field FF to detect failure of the tensor criterion.

Source excerpts

“Define the Q-linear map Φ : F ⊗Q F → F by Φ(a⊗b) = (∂a)(∂b). For every P ∈ Mat3(B), the product rule and (31) give ∑ i βiΦ ( (pi ⊗ 1)TP(1 ⊗ pi) ) = 4 ∑ 2 α=1 mF (Pαα).” · Show in paper