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

Short Egyptian fractions

There are absolute constants c1,c2>0c_1, c_2 > 0 and b0b_0 such that, for every integer b≥b0b \ge b_0, c1log⁡log⁡b≤N(b)≤c2log⁡log⁡bc_1 \log\log b \le N(b) \le c_2 \log\log b, and in particular the upper bound holds for every integer numerator 1≤a<b1 \le a < b.

There are absolute constants c,C>0c, C > 0 and k0k_0 such that, for every integer k≥k0k \ge k_0, ck≤log⁡log⁡F(k)≤Ckck \le \log\log F(k) \le Ck, where F(k)F(k) is the number of representations of 11 as a sum of kk distinct positive unit fractions.

For every sufficiently large integer kk, exp⁡(exp⁡(k/600))≤v(k)≤1+k2k−1\exp(\exp(k/600)) \le v(k) \le 1 + k^{2^{k-1}}, and more precisely log⁡2257≤lim inf⁡k→∞log⁡log⁡v(k)k≤lim sup⁡k→∞log⁡log⁡v(k)k≤log⁡2\frac{\log 2}{257} \le \liminf_{k\to\infty} \frac{\log\log v(k)}{k} \le \limsup_{k\to\infty} \frac{\log\log v(k)}{k} \le \log 2, so log⁡log⁡v(k)=Θ(k)\log\log v(k) = \Theta(k).

For every ε>0\varepsilon > 0 there is an integer mεm_\varepsilon such that every integer m≥mεm \ge m_\varepsilon occurs as an exact denominator in a representation of 11 by distinct positive unit fractions with at most (257log⁡2+ε)log⁡log⁡m(\frac{257}{\log 2} + \varepsilon)\log\log m terms.

There are absolute constants DM>1D_M > 1, L>0L > 0, and S0>1S_0 > 1 (with DX=DM+2D_X = D_M + 2 and DC=4DXD_C = 4D_X) such that for every real S≥S0S \ge S_0 and every integer C∈[eDCS,e2DCS]C \in [e^{D_C S}, e^{2D_C S}], there are an integer M∈(eS,eDMS]M \in (e^S, e^{D_M S}] and a set G⊆{1,…,⌊X⌋}G \subseteq \{1, \dots, \lfloor X \rfloor\} (X=eDXSX = e^{D_X S}) with ∣{1,…,⌊X⌋}∖G∣≤X/8|\{1, \dots, \lfloor X \rfloor\} \setminus G| \le X/8, such that for each u∈Gu \in G, u/(MC)u/(MC) is a sum of at most Llog⁡SL \log S unit fractions with integer denominators at least 22 (repetitions allowed).

For every fixed pair of real numbers D,r≥1D, r \ge 1, there is S0(D,r)S_0(D, r) such that for S≥S0(D,r)S \ge S_0(D, r) and any X,Y∈RX, Y \in \mathbb{R}, N∈Z≥0N \in \mathbb{Z}_{\ge 0} with S2log⁡S≤log⁡X≤DS\frac{S}{2\log S} \le \log X \le DS, X1/2≤Y≤XX^{1/2} \le Y \le X, and N≤eDSN \le e^{DS}, one has ∑1≤h≤YdX(N+h)r≤Yexp⁡(S1/4)\sum_{1 \le h \le Y} d_X(N+h)^r \le Y \exp(S^{1/4}).

Methods

Reducing the uniform Egyptian fraction problem for all numerators 1≤a<b1 \le a < b to a dense set G⊆{1,…,⌊X⌋}G \subseteq \{1, \dots, \lfloor X \rfloor\} of short expansions over MCMC by scaling n=gAM∈[X/2,X]n = gAM \in [X/2, X] and writing n=u+(n−u)n = u + (n-u) with u,n−u∈Gu, n-u \in G.

Source excerpts

“Write H = {1, . . . , ⌊X⌋} \ G. Among the n − 1 positive splits n = u + (n − u), at most 2|H| ≤ X/4 have an entry outside G. Since n − 1 ≥ X/2 − 1 > X/4 for large S, some split has both entries in G. Their expansions together express n/(MC) = gA/C using at most 2Llog S unit fractions.” · Show in paper

Controlling the distribution of residues −Cjt(modu)-C_j t \pmod u for divisors t∣Mt \mid M using the Erdős–Turán discrepancy inequality, van der Corput differencing for the deterministic prime-product list T0T_0, and random prime-product lists T1,…,TRT_1, \dots, T_R with additive-character orthogonality.

Source excerpts

“We use the Erdős–Turán discrepancy inequality [9, Part I, Theorem III].” · Show in paper

“For the smaller moduli, two independent products of sampled primes supply the cancellation. Their integer values are smaller than a suitable divisor of the modulus, so unique factorization controls their residue distributions.” · Show in paper

Propagating exceptional sets backwards through O(log⁡S)O(\log S) levels by combining a uniform rr-th moment bound for truncated divisor counts dX(N+h)d_X(N+h) with Hölder's inequality.

Source excerpts

“Thus Lemma 3.1, with the fixed parameters D∗, r, applies uniformly at every level and to every list entry. Hölder’s inequality gives ∑ h∈Hj+1 dX(N + h) ≤ |Hj+1|α ( ∑ 1≤h≤Y dX(N + h)r )1/r ≤ Y eS1/4/r δαj+1.” · Show in paper

Reserving 1/m1/m and running an unreduced greedy prefix that skips denominator mm until the remainder R/qR/q is smaller than 1/m1/m and every prefix reciprocal, while retaining divisors of qq at every multiplicative scale.

Source excerpts

“We instead reserve 1/m and use a greedy construction that skips denominator m. It produces a distinct prefix and makes the remaining sum smaller than every reserved reciprocal. Any positive expansion of this remainder then avoids both the marker and the prefix denominators.” · Show in paper

Constructing a common integer KmK_m whose divisors represent every integer in [1,m4][1, m^4] as a sum of at most 1616 rationals e/te/t with e∣Kme \mid K_m, using Gallagher's larger sieve, Vinogradov's three-prime theorem, and bilinear estimates in Fp\mathbb{F}_p.

Source excerpts

“We estimate exceptional primes by counting congruent pairs, the mechanism underlying Gallagher’s larger sieve [10].” · Show in paper

“We use the quantitative form of Vinogradov’s three-prime theorem: there is an absolute c > 0 such that every sufficiently large odd integer u has at least cu2/(log u)3 ordered representations as a sum of three primes” · Show in paper