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

The Falconer distance conjecture in all dimensions

For every integer d≥2d \ge 2 and every compact subset E⊂RdE \subset \mathbb{R}^d with dim⁡HE>d/2\dim_H E > d/2, the Euclidean distance set Δ(E)={∣x−y∣:x,y∈E}\Delta(E) = \{|x - y| : x, y \in E\} has positive Lebesgue measure, L1(Δ(E))>0\mathcal{L}^1(\Delta(E)) > 0.

For fixed k≥2k \ge 2, k−1≤q≤k−1/2<t<kk - 1 \le q \le k - 1/2 < t < k, and tube-width constant CC (and ζ>0\zeta > 0 when q=k−1q = k - 1), there exist ϵ,c>0\epsilon, c > 0 such that for all sufficiently small δ\delta, any bounded point subprobability μ\mu in Rk+1\mathbb{R}^{k+1} and direction subprobability ν\nu on SkS^k with tt- and qq-ball bounds of loss δ−ϵ\delta^{-\epsilon} down to δ\delta (plus the hyperplane-plate bound ν{w:dist⁡(w,H)≤r}≤δ−ϵrζ\nu\{w : \operatorname{dist}(w, H) \le r\} \le \delta^{-\epsilon} r^\zeta when q=k−1q = k - 1) satisfy (μ⊗ν)(E)≤δc(\mu \otimes \nu)(E) \le \delta^c whenever each direction fiber EwE_w of a measurable incidence relation E⊂Rk+1×SkE \subset \mathbb{R}^{k+1} \times S^k is covered by at most δ−q−ϵ\delta^{-q-\epsilon} tubes parallel to ww of width CδC\delta.

Let k≥2k \ge 2 and k−1/2<s<kk - 1/2 < s < k, and let ρ,ν\rho, \nu be bounded, separated ss-Frostman probabilities in Rk+1\mathbb{R}^{k+1} whose directional measures on GNG_N have plate bounds CuζC u^\zeta uniformly for 2−N≤u≤12^{-N} \le u \le 1 with fixed ζ>0\zeta > 0. If k−1k - 1 is available for GNG_N, then k−1/2k - 1/2 is available.

For 1≤H≤R1 \le H \le R, K,L≥1K, L \ge 1, and S=d/2S = d/2, if α\alpha is a probability in a fixed bounded subset of Rd\mathbb{R}^d with α(B(x,r))≤KrS\alpha(B(x,r)) \le K r^S for H−1≤r≤1H^{-1} \le r \le 1 and ξ\xi is a subprobability on Sd−1S^{d-1} with ξ(B(w,r))≤LrS\xi(B(w,r)) \le L r^S for R−1≤r≤1R^{-1} \le r \le 1, then ∫−HH∫Sd−1∣α^(sw)∣2 dξ(w) ds≤CLK(1+log⁡H)\int_{-H}^H \int_{S^{d-1}} |\widehat{\alpha}(sw)|^2 \, d\xi(w) \, ds \le C L K (1 + \log H).

Under the depth conditions 0≤a<p≤v≤N0 \le a < p \le v \le N, p−a≤v−p+2σNp - a \le v - p + 2\sigma N, L≤v−p+2σNL \le v - p + 2\sigma N, v−p≥10σNv - p \ge 10\sigma N, and the parallel test [gj,p][g_j, p] with a≤gj<pa \le g_j < p and gj≤a+2σNg_j \le a + 2\sigma N imposed by the admissible endpoint symbols, the general independent angular energy satisfies HX,Y(a,v)≤R−∑j=12(Aj,p−Aj,a)+min⁡j=1,2kj(a,p)+Od(σ)+o(1)max⁡P⊂X,Q⊂YHP,Q(p,v)+O(R−T)\mathsf{H}^{X,Y}(a,v) \le R^{-\sum_{j=1}^2 (A_{j,p} - A_{j,a}) + \min_{j=1,2} k_j(a,p) + O_d(\sigma) + o(1)} \max_{P \subset X, Q \subset Y} \mathsf{H}^{P,Q}(p,v) + O(R^{-T}) for every fixed T>0T > 0.

Let (αN)N≥K(\alpha_N)_{N \ge K} be finite positive Borel measures on R\mathbb{R} with αN+1≤αN\alpha_{N+1} \le \alpha_N and α∞=lim⁡NαN\alpha_\infty = \lim_N \alpha_N. If τN\tau_N are finite complex Borel measures satisfying ∑N≥K∥αN−τN∥TV<∞\sum_{N \ge K} \|\alpha_N - \tau_N\|_{\mathrm{TV}} < \infty and ∑N≥K(∫2N−1≤∣r∣≤2N+1∣τN^(r)∣2 dr)1/2<∞\sum_{N \ge K} \left(\int_{2^{N-1} \le |r| \le 2^{N+1}} |\widehat{\tau_N}(r)|^2 \, dr\right)^{1/2} < \infty, then α∞\alpha_\infty is absolutely continuous with respect to Lebesgue measure.

For the starting pair in Proposition 5.10, the fixed parameters can be chosen in the order β,κ,σ,ω\beta, \kappa, \sigma, \omega so that some γ>0\gamma > 0 satisfies ∫2N−1≤∣r∣≤2N+1∣τN^(r)∣2 dr≲2−γN\int_{2^{N-1} \le |r| \le 2^{N+1}} |\widehat{\tau_N}(r)|^2 \, dr \lesssim 2^{-\gamma N} for all sufficiently large NN.

Methods

Regularization of a probability measure on a unit dyadic cube into disjoint regular classes with equi-Lipschitz mass profiles mαm_\alpha and density pruning/multiplication along selected dyadic depths.

Source excerpts

“Lemma 3.1 (Regular classes). Let n ≥ 1, let ρ be a probability on a unit dyadic cube in R n , and put R = 2N . For every fixed 0 < λ < 1 and all sufficiently large N, there are pairwise disjoint sets Eα, each a union of depth-N cubes...” · Show in paper

“Lemma 3.2 (Density pruning and multiplication). Let ρ be a source probability and ν a finite pin measure, both on R n , and let 0 ≤ a0 < · · · < aJ ≤ N be a fixed finite collection of dyadic depths.” · Show in paper

Slicing incidence relations and Frostman/cap-bounded measures to Haar-random codimension-hh subspaces with width-δ\delta bands, preserving ball bounds, active tube covers, and subpower incidence mass.

Source excerpts

“Lemma 4.5 (Selection of slices). Let 0 ≤ h ≤ n − 2. Let point and direction subprobabilities µ, ν in R n have t- and q-ball bounds with subpower losses, where t, q > h.” · Show in paper

Global stationary-phase direct and inverse expansions on the sphere Sn∗S^{n_*} using global differential operators LjL_j and BjB_j built from the spherical Laplacian ΔSn∗\Delta_{S^{n_*}}, combined with separated spatial Fourier series.

Source excerpts

“Lemma 7.5 (Direct and inverse sphere expansions).” · Show in paper

“Lemma 7.6 (Separated spatial expansion).” · Show in paper

A two-depth potential V(J)=min⁡a≤i<j≤t, j−i≥10q0(Ai+Aj)\mathcal{V}(J) = \min_{a \le i < j \le t, \, j - i \ge 10q_0} (A_i + A_j) on depth intervals to pay for positive step costs in recursive reductions between positive distance states and angular energy states.

Source excerpts

“Definition 10.1 (Two-depth potential). For an integer interval J = [a, t] of width at least 10q0, put V(J) = min a≤i<j≤t j−i≥10q0 (Ai + Aj ). (10.1)” · Show in paper

Monotone-measure shell reconstruction using smooth dyadic Fourier cutoffs PN=QN−QN−1P_N = Q_N - Q_{N-1} and summable positive increments ζn=αn−αn+1\zeta_n = \alpha_n - \alpha_{n+1} to prove absolute continuity of a decreasing limit of positive measures without a quantitative convergence rate.

Source excerpts

“Lemma 11.4 (Monotone measure shell criterion).” · Show in paper