Skip to content
#small language model Open access

Continued-Fraction Statistics Cannot Tell Known Transcendentals from Unknown Ones or from a Cubic Irrational ── the geometric mean of partial quotients, the growth of denominators and the share of ones all fall within the scatter of random numbers for eight named numbers, and only periodic quadratic numbers and the patterned e stand out ── the separator is whether one speaks of almost all numbers in measure or of this number by name ── [Paper 1021]

Sep 2026 · Zenodo (CERN European Organization for Nuclear Research)

Abstract

Whether π + e, πe or γ is transcendental is not known. On the other hand, it is known that for almost all real numbers the geometric mean of the partial quotients of the continued fraction approaches Khinchin’s constant (Khinchin). This paper expands to 3000 terms the continued fractions of numbers known to be transcendental, numbers not known to be, and a cubic algebraic number, and lines them up by the same statistics. No new theorem or law is claimed, and no number is proved transcendental. Scope of this paper (scope note): No new theorem or law is claimed──that for almost all reals the geometric mean of partial quotients approaches Khinchin’s constant (Khinchin), Lévy’s constant for the growth of denominators, the Gauss-Kuzmin distribution of partial quotients, that continued fractions of quadratic irrationals are periodic, the pattern of the continued fraction of e, the transcendence of π (Lindemann), of e to the power π (Gelfond) and of ln 2 and ln π, that the transcendence of π + e, πe and γ is open, and the computations on Khinchin’s constant and its continued fractions (Bailey, Borwein and Crandall, 1997) are all known. No number is proved transcendental──statistics of 3000 terms prove nothing about transcendence or about being typical. The number at the edge is not read as evidence──that π + e sits at an edge is treated as the ordinary scatter of lining up eight numbers. Whether the partial quotients of the cube root of 2 are bounded is not entered──this is open, and this paper only looked at 3000 terms. Relation to earlier papers: the scope note of Paper 811 said that it does not enter transcendence and does not use that π and e are transcendental; this paper sits in that seat and lines up known and undecided transcendentals by continued-fraction statistics (no earlier occupant). Paper 812 showed that what decides how easily a number is approximated is whether its partial quotients are bounded──here it is counted that only the bounded periodic numbers and the growing e stand out, while the rest fall into the same scatter. Paper 176 showed that "small" has two senses, small in measure and small in category, which do not agree──here it is counted that Khinchin’s theorem is "almost all" in the sense of measure and says nothing about each named number. Paper 804 showed that measured heights alone cannot rule out fractions of e──here it is counted that continued-fraction statistics alone cannot separate transcendental from algebraic. What is added is lining up eight named numbers by the same statistics, giving their positions within the scatter of 100 random numbers, treating π + e at the edge together with the correlation between the quantities, and showing that only numbers with a period or a fixed pattern stand out. First, and this is the core. The eight named numbers cannot be told apart within the scatter of random numbers──the geometric means of partial quotients were 2.667237, 2.699672, 2.659773 and 2.623448 for the known transcendentals π, e to the power π, ln 2 and ln π, 2.600972, 2.677830 and 2.684892 for the undecided π + e, πe and γ, and 2.630640 for the algebraic cube root of 2. One hundred random numbers had mean 2.680329 and standard deviation 0.055238, and the eight positions fell within 1.436645 standard deviations (Sections 2 and 3). Second, all three statistics sit around the same constants──Khinchin’s constant is 2.685452001, Lévy’s constant for the growth of denominators is 1.186569110, and the share of partial quotients equal to 1 tends to log₂(4/3) = 0.415037499; the values for the eight numbers scatter around these (Section 2). Third, even π + e, at the edge, is no evidence──the geometric mean of π + e lies 6 percent from the bottom of the 100 random numbers (−1.436645 standard deviations), and its share of ones, 0.437146, slightly exceeds the random maximum 0.435478. The three quantities move together (correlation -0.750915 between geometric mean and share of ones), and among eight numbers it is ordinary for one to sit at an edge (Section 3). Fourth, only numbers with a period or a fixed pattern stand out──√2 has geometric mean 2 and the golden ratio 1, while e keeps growing, 2.592229 at 60 terms and 5.325766 at 600 terms, beyond the top of the random scatter (Section 4). Fifth, a cubic algebraic number also looks like a typical number──unlike periodic quadratic numbers, the cube root of 2 has geometric mean 2.630640 (-0.899555 standard deviations among the random numbers) and stands among the transcendentals (Section 3). Sixth, the separator is whether one speaks of almost all numbers in measure or of this number by name──Khinchin’s theorem speaks of almost all numbers and does not say which are exceptions. Which side a named number falls on is not settled by numerics (Section 5). Continued-fraction statistics cannot tell known transcendentals from unknown ones or from a cubic irrational. Expanding eight numbers to 3000 terms, π, e to the power π, ln 2 and ln π, π + e, πe and γ, and the cube root of 2, the geometric means of partial quotients lay between 2.600972 and 2.699672, within 1.436645 standard deviations of the 100 random numbers, whose mean was 2.680329 and standard deviation 0.055238──even π + e at the edge is within the ordinary range, as a single departure in which the three quantities move together. Only the periodic √2 and golden ratio and the patterned e (5.325766 at 600 terms) stand out. The separator is whether one speaks of almost all numbers in measure or of this number by name. Placed among the earlier papers──Paper 812 separated ease of approximation by bounded and unbounded partial quotients, Paper 176 separated "small" into measure and category, and Paper 804 showed that heights cannot rule out fractions of e. Here it was counted that continued-fraction statistics do not distinguish the transcendence that Paper 811 did not enter. To be honest──nothing new is known about which numbers are transcendental. On the making of this work: The ideas and content of this work stem from the author's own considerations. Assistance from an AI (a large language model) was used for structuring, English translation, and checking the algebra. Any remaining errors or misinterpretations are solely the author's. Feedback and corrections are sincerely appreciated. Keywords: continued fraction, Khinchin's constant, Lévy's constant, Gauss-Kuzmin distribution, transcendental number, algebraic number. ----- π+e や πe や γ が超越数かどうかは分かっていない。一方、ほとんどすべての実数では、連分数の部分商の幾何平均がヒンチンの定数に近づくことが知られている(ヒンチン)。本稿は、超越と分かっている数・分かっていない数・三次の代数的数の連分数を 3000 項ずつ展開し、同じ統計で並べる。新しい定理も法則も主張せず、どの数が超越かも証明しない。 本稿の射程(射程注記):新しい定理も法則も主張しない──ほとんどすべての実数で部分商の幾何平均がヒンチンの定数に近づくこと(ヒンチン)、分母の伸びのレヴィの定数、部分商の分布(ガウス=クズミンの分布)、二次の無理数の連分数が周期的であること、e の連分数の形、π の超越性(リンデマン)、e の π 乗の超越性(ゲルフォント)、ln 2 と ln π の超越性、π+e・πe・γ の超越性が未解決であること、ヒンチンの定数とその連分数についての計算(ベイリー、ボーウェイン、クランドール、1997)は、いずれも既知である。どの数が超越かを証明しない──3000 項の統計は、超越性についても、典型的かどうかについても何も証明しない。端の数を証拠と読まない──π+e が端に来たことは、八つを並べたときの普通の揺れとして扱う。2 の三乗根の部分商が有界かどうかに入らない──これは未解決で、本稿は 3000 項を見ただけである。既刊との関係:論文811 の射程注記は、超越性に入らない、π や e が超越数であることは使わないと書いた──本稿はその席に座り、超越と分かっている数と分かっていない数を連分数の統計で並べる(先客なし)。論文812 は、近似のしやすさを決めるのが部分商の有界か非有界かであることを示した──ここでは、有界な周期を持つ数と増え続ける e だけが外れ、ほかは同じ散らばりに入ることを数える。論文176 は、「小さい」には測度で小さいことと範疇で小さいことの二つがあり、両立しないことを示した──ここでは、ヒンチンの定理が測度の意味の「ほとんどすべて」であり、名前のある数の一つ一つには何も言わないことを数える。論文804 は、測った高さだけでは e の分数を否定できないことを示した──ここでは、連分数の統計だけでは超越か代数的かを分けられないことを数える。加えたのは、名前のある八つの数を同じ統計で並べたこと、乱数 100 個の散らばりの中での位置を出したこと、端に来た π+e を、量どうしの相関とともに扱ったこと、外れる数が周期と決まった形の数だけであることを示したことである。 第一に、これが本稿の芯である。名前のある八つの数は、乱数の散らばりの中で見分けがつかない──部分商の幾何平均は、超越と分かっている π・e の π 乗・ln 2・ln π で 2.667237・2.699672・2.659773・2.623448、分かっていない π+e・πe・γ で 2.600972・2.677830・2.684892、代数的な 2 の三乗根で 2.630640 だった。乱数の数 100 個は平均 2.680329・標準偏差 0.055238 で、八つの位置は標準偏差の 1.436645 倍以内に収まった(第2節・第3節)。 第二に、三つの統計はどれも同じ定数のまわりにある──ヒンチンの定数は 2.685452001、分母の伸びのレヴィの定数は 1.186569110、部分商が 1 になる割合は log₂(4/3)=0.415037499 で、八つの数の値はそのまわりに散らばる(第2節)。 第三に、いちばん端の π+e も、証拠にはならない──π+e の幾何平均は乱数 100 個の中で下から 6 パーセントの位置(標準偏差の −1.436645 倍)で、1 の割合 0.437146 は乱数の最大 0.435478 をわずかに超えた。三つの量は一緒に動き(幾何平均と 1 の割合の相関 -0.750915)、八つを並べれば一つが端に来るのはふつうである(第3節)。 第四に、外れるのは、周期を持つ数と形の決まった数だけである──√2 の幾何平均は 2、黄金比は 1 で、e は 60 項で 2.592229、600 項で 5.325766 と増え続け、乱数の散らばりの上限を超える(第4節)。 第五に、三次の代数的数も「典型的な数」に見える──2 の三乗根は、周期を持つ二次の数と違い、幾何平均 2.630640(乱数の中で標準偏差の -0.899555 倍)で超越数と並ぶ(第3節)。 第六に、分離子は、測度で言う「ほとんどすべて」か、名前で指す「この数」か──ヒンチンの定理はほとんどすべての数について言い、どの数が例外かは言わない。名前のある数がどちらに入るかは、数値では決まらない(第5節)。 連分数の統計は、超越と分かっている数と、分かっていない数と、三次の代数的数を見分けない。π・e の π 乗・ln 2・ln π、π+e・πe・γ、2 の三乗根の八つを 3000 項ずつ展開すると、部分商の幾何平均は 2.600972 から 2.699672 の間にあり、乱数 100 個の平均 2.680329・標準偏差 0.055238 に対して、位置は 1.436645 倍以内だった──端の π+e も、三つの量が一緒に動く一つの揺れとして普通の範囲にある。外れるのは周期を持つ √2 と黄金比、形の決まった e(600 項で 5.325766)だけである。分離子は、測度で言う「ほとんどすべて」か、名前で指す「この数」かである。既刊との位置──論文812 は近似のしやすさを部分商の有界と非有界で分け、論文176 は「小さい」を測度と範疇に分け、論文804 は高さでは e の分数を否定できないことを示した。ここでは、論文811 が入らなかった超越性を、連分数の統計が見分けないことを数えた。正直に言えば──どの数が超越かについて、何一つ新しくは分かっていない。 作成にあたって:本稿の着想と内容は、著者自身の考察に基づくものです。文章の構成整理や英訳、数式の確認には AI(大規模言語モデル)の助力を得ました。最終的な内容の解釈や誤りがあれば、それらはすべて著者の責に帰します。お気づきの点があれば、ご教示いただければ幸いです。 キーワード:連分数、ヒンチンの定数、レヴィの定数、ガウス=クズミン、超越数、代数的数。

View source

Similar papers

#small language model Dataset Open access Oct 2026

Socratic guiding questions in synthetic arithmetic data: matched LoRA runs (revision v2)

Supporting data, adapters, predictions and code for the article *Low-Cost LoRA Fine-Tuning of Small Language Models for Multi-Step Arithmetic Reasoning* by Jake O'Grady, Asena Isik Gürhan, Chee Fong Ting and Effirul Ramlan (University of Galway). We generated 20,000 GSM8K-derived arithmetic problems with step-by-step s...

O'Grady, Jake, Gürhan, Asena Isik, Chee, Fong Ting et al. · 465 citations
#computer vision Open access Jun 2016

Software Development in Startup Companies: The Greenfield Startup Model

The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.

Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al. · 178 citations · ⚡14
#computer vision Open access Oct 2016

Software Startups - A Research Agenda

Software startup companies develop innovative, software-intensive products within limited timeframes and with few resources, searching for sustainable and scalable business models.

M. Unterkalmsteiner, P. Abrahamsson, Xiaofeng Wang et al. · 157 citations · ⚡17
#machine learning Review Open access Oct 2016

“Failures” to be celebrated: an analysis of major pivots of software startups

This study conducts a case survey study based on the secondary data of the major pivots happened in 49 software startups, and demonstrates that customer need pivot is the most common among all pivot types.

Sohaib Shahid Bajwa, Xiaofeng Wang, Anh Nguyen-Duc et al. · 127 citations · ⚡15
#computer vision Review Open access May 2015

A survey study on major technical barriers affecting the decision to adopt cloud services

The comparison of adopter and non-adopter sample reveals three potential adoption inhibitor, security, data privacy, and portability, which underlines the importance of the technical and security perspectives for research investigating the adoption of technology.

Nattakarn Phaphoom, Xiaofeng Wang, S. Samuel et al. · 111 citations · ⚡8
#computer vision Open access Feb 2018

Lean Internal Startups for Software Product Innovation in Large Companies: Enablers and Inhibitors

This study investigates how Lean internal startup facilitates software product innovation in large companies and identifies its enablers and inhibitors, and shows the potential of the method-in-action framework to investigate the Lean startup approach in non-startup context.

Henry Edison, Nina M. Smørsgård, Xiaofeng Wang et al. · 78 citations · ⚡6

Related blog posts

Microsoft Research Blog Oct 12, 2026

Simulating student learning for better AI tutoring

StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics. The post Simulating student learning for better AI tutoring appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.