arxiv: 2603.21852 · v2 · submitted 2026-03-23 · 💻 cs.SC · cs.LG

Recognition: 2 theorem links

· Lean Theorem

All elementary functions from a single binary operator

Andrzej Odrzywo{\l}ek

Authors on Pith no claims yet

Pith reviewed 2026-05-15 01:24 UTC · model grok-4.3

classification 💻 cs.SC cs.LG

keywords elementary functionsbinary operatorEML operatorfunctional completenesssymbolic regressionscientific calculatorexp minus log

0 comments

The pith

One binary operator together with the constant 1 generates every elementary function on a scientific calculator.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that the binary operator eml(x, y) defined as exp(x) minus the natural log of y, when combined only with the number 1, can construct all standard elementary functions. This covers arithmetic operations, exponentiation, roots, logarithms, trigonometric functions, and constants such as e, pi, and the imaginary unit i. A reader might care because it reveals a hidden simplicity in continuous mathematics, where a single uniform operation suffices instead of many separate functions. The uniform binary tree structure of expressions also supports training optimizers to discover exact formulas from numerical data.

Core claim

A single two-input operator eml(x,y)=exp(x)-ln(y), together with the constant 1, generates the standard repertoire of a scientific calculator including constants e, pi, and i as well as all arithmetic, algebraic, and transcendental elementary functions. Every such function is obtained by explicit nesting, for instance exp(x) equals eml(x,1) and ln(x) equals eml(1, eml(eml(1,x),1)). The operator was identified by exhaustive search and verified constructively for the calculator basis.

What carries the argument

The EML binary operator eml(x,y)=exp(x)-ln(y) that acts as the sole primitive, allowing all other operations to be built by composition in binary trees with leaves equal to 1.

If this is right

Every elementary function reduces to a binary tree composed entirely of identical eml nodes.
Constants like e, pi, and i emerge from finite nestings starting from 1.
Gradient-based optimization on EML trees can recover exact closed-form expressions from data when the underlying function is elementary.
Arbitrary data can be fitted, but exact formula recovery occurs precisely when the law is elementary.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If this operator works without branch issues, it might allow single-gate hardware for mathematical computations analogous to NAND gates in logic.
The approach could extend symbolic regression methods to deeper trees or other function classes.
Proving identities or equivalences in elementary functions might simplify under this uniform representation.

Load-bearing premise

That finite explicit nestings of the eml operator applied only to the constant 1 can produce every elementary function without hidden restrictions from branch cuts in complex logarithms or domain issues.

What would settle it

A counterexample consisting of an elementary function such as arcsin(x) or the constant pi that cannot be matched by any finite eml nesting starting from 1, or numerical mismatch due to branch cut choices in the constructed expressions.

Figures

Figures reproduced from arXiv: 2603.21852 by Andrzej Odrzywo{\l}ek.

**Figure 2.** Figure 2: Examples of binary EML trees equivalent to few important simplest formulas. [PITH_FULL_IMAGE:figures/full_fig_p023_2.png] view at source ↗

read the original abstract

A single two-input gate suffices for all of Boolean logic in digital hardware. No comparable primitive has been known for continuous mathematics: computing elementary functions such as sin, cos, sqrt, and log has always required multiple distinct operations. Here I show that a single binary operator, eml(x,y)=exp(x)-ln(y), together with the constant 1, generates the standard repertoire of a scientific calculator. This includes constants such as e, pi, and i; arithmetic operations including addition, subtraction, multiplication, division, and exponentiation as well as the usual transcendental and algebraic functions. For example, exp(x)=eml(x,1), ln(x)=eml(1,eml(eml(1,x),1)), and likewise for all other operations. That such an operator exists was not anticipated; I found it by systematic exhaustive search and established constructively that it suffices for the concrete scientific-calculator basis. In EML (Exp-Minus-Log) form, every such expression becomes a binary tree of identical nodes, yielding a grammar as simple as S -> 1 | eml(S,S). This uniform structure also enables gradient-based symbolic regression: using EML trees as trainable circuits with standard optimizers (Adam), I demonstrate the feasibility of exact recovery of closed-form elementary functions from numerical data at shallow tree depths up to 4. The same architecture can fit arbitrary data, but when the generating law is elementary, it may recover the exact formula.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

A single binary operator eml(x,y)=exp(x)-ln(y) with constant 1 can generate the elementary functions via nesting, found by search, but the complex cases rest on unshown branch choices.

read the letter

The core result is that this specific operator, discovered through exhaustive search, lets you build exp, ln, and then the rest of the calculator functions as finite trees with leaves fixed at 1. The explicit rewrites for exp(x) = eml(x,1) and ln(x) = eml(1, eml(eml(1,x),1)) are direct and non-circular, which is the clean part of the work. Treating the same trees as differentiable circuits for symbolic regression is a practical follow-on that can recover exact formulas at small depths when the target is elementary. That combination of minimal basis plus regression demo is what stands out as new. The constructions stay syntactic and use only standard properties, so no hidden fitting or self-reference appears in the real case. The soft spot is the extension to pi, i, sin and the complex plane. Ln is multi-valued, so any nesting that produces -1 or i selects a sheet; the abstract asserts sufficiency but does not display the actual trees or a consistent branch rule that matches principal values on open sets. If those trees exist and stay on the right sheet, the claim holds; without them visible, the complex part remains plausible rather than demonstrated. This is the kind of paper that belongs in a computer-algebra or symbolic-computation venue. Readers who care about minimal operator sets or gradient-based formula discovery will find the constructions and the regression experiment useful. It is coherent on its own terms and shows clear thinking about the basis question, so it deserves a serious referee who can ask for the missing trees and a short branch discussion. I would send it to review rather than desk-reject.

Referee Report

2 major / 2 minor

Summary. The paper claims that the binary operator eml(x,y)=exp(x)-ln(y), together with the constant 1, generates all elementary functions (including exp, ln, arithmetic operations, exponentiation, and transcendentals such as sin) as well as the constants e, pi, and i. Explicit constructions are supplied for exp(x)=eml(x,1) and ln(x)=eml(1,eml(eml(1,x),1)); the remainder are asserted to follow by similar finite nestings. The paper further shows that EML trees can be used as trainable circuits for gradient-based symbolic regression, recovering exact elementary formulas from data at shallow depths.

Significance. If the central claim is fully substantiated, the result supplies a uniform, parameter-free basis for the entire elementary-function repertoire, analogous to a single gate for Boolean logic. The constructive syntactic rewrites, the simple grammar S -> 1 | eml(S,S), and the demonstration of exact recovery via Adam optimization on EML trees constitute clear strengths for symbolic computation and automated formula discovery.

major comments (2)

[Abstract] Abstract: the assertion that the remainder of the elementary functions 'follow likewise' is not supported by any explicit eml-tree; without concrete expressions for pi, i, sin, or the arithmetic operations, the claim that the operator generates the full scientific-calculator basis remains unverified.
[Main text (constructions)] Main constructions: the paper does not address branch selection for the multi-valued complex logarithm. Any tree that produces i or pi (via ln(-1)) or that implements sin(z)=(e^{iz}-e^{-iz})/(2i) necessarily selects a particular sheet; without an explicit rule guaranteeing the principal branch, the generated functions may differ from the standard definitions on open sets.

minor comments (2)

The grammar S -> 1 | eml(S,S) is stated clearly but would benefit from a small diagram showing the tree for a non-trivial function such as addition or multiplication.
The symbolic-regression experiments report recovery at depths up to 4; the precise success rates and the loss function used should be tabulated for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The comments highlight two areas where the manuscript can be strengthened: providing explicit EML trees for additional functions and clarifying branch selection for the complex logarithm. We address each point below and will incorporate revisions to improve clarity and verifiability.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion that the remainder of the elementary functions 'follow likewise' is not supported by any explicit eml-tree; without concrete expressions for pi, i, sin, or the arithmetic operations, the claim that the operator generates the full scientific-calculator basis remains unverified.

Authors: We agree that the abstract and main text provide only illustrative examples (exp and ln) and assert that the remainder follow by similar finite nestings without listing them explicitly. While the grammar S -> 1 | eml(S,S) and the constructive existence proof establish that such trees exist, explicit expressions would make the claim easier to verify. We will add a new subsection with concrete EML trees for arithmetic operations (+, -, *, /, ^), the constants e, pi, i, and the functions sin, cos, and sqrt in the revised manuscript. revision: yes
Referee: [Main text (constructions)] Main constructions: the paper does not address branch selection for the multi-valued complex logarithm. Any tree that produces i or pi (via ln(-1)) or that implements sin(z)=(e^{iz}-e^{-iz})/(2i) necessarily selects a particular sheet; without an explicit rule guaranteeing the principal branch, the generated functions may differ from the standard definitions on open sets.

Authors: The constructions implicitly use the principal branch of the logarithm, consistent with standard definitions of elementary functions in the complex domain. However, the manuscript does not state this explicitly or discuss the domains on which the identities hold. We will add a dedicated paragraph in the constructions section specifying that all logarithms are taken on the principal branch (with the usual branch cut along the negative real axis) and note the open sets where the generated functions coincide with the standard ones. revision: yes

Circularity Check

0 steps flagged

No circularity: explicit constructive rewrites from operator definition

full rationale

The paper defines eml(x,y) := exp(x) - ln(y) and supplies finite nestings with the constant 1 that algebraically recover exp, ln, and the remaining elementary operations. Each identity is obtained by direct substitution and cancellation (e.g., the three-level expression for ln(x) evaluates to ln(x) by the functional equation of exp and ln). These steps are self-contained syntactic derivations; they do not invoke fitted parameters, rename known results, or rest on self-citations. The mention of exhaustive search is only a discovery narrative and does not appear in any load-bearing derivation. The claimed universality therefore rests on independent algebraic verification rather than circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The claim rests on the standard analytic properties of exp and ln (including their behavior over the complex numbers) and introduces the eml operator as the sole new primitive; no numerical parameters are fitted.

axioms (1)

standard math Standard properties of the exponential and natural logarithm functions, including inverses and extension to complex numbers.
The operator is defined directly in terms of exp and ln, so all constructions inherit these background facts.

invented entities (1)

eml binary operator no independent evidence
purpose: Single primitive sufficient to generate all elementary functions and arithmetic operations.
Defined as exp(x) - ln(y); its universality is the central claim to be verified by construction.

pith-pipeline@v0.9.0 · 5559 in / 1341 out tokens · 54326 ms · 2026-05-15T01:24:11.103189+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel echoes

?

echoes
ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

a single binary operator, eml(x,y)=exp(x)-ln(y), together with the constant 1, generates ... constants such as e, pi, and i; arithmetic operations ... transcendental and algebraic functions
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction echoes

?

echoes
ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

S→1|eml(S,S) ... uniform structure ... binary tree of identical nodes

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Additive Atomic Forests for Symbolic Function and Antiderivative Discovery
cs.LG 2026-05 unverdicted novelty 7.0

A derivative algebra with EML and SOL primitives plus additive atomic forests enables simultaneous symbolic recovery of functions and antiderivatives from data, matching or exceeding XGBoost on 13 of 17 benchmarks wit...
Non-Monotone Response Modules and Cascades from the EML Operator for Reduced Models of Biological Dynamics
math.DS 2026-05 unverdicted novelty 6.0

EML operator serves as a grammar to generate single-block activation-suppression modules that produce non-monotone dynamics in reduced biological ODEs, matching data on PKA-R and Rho-GTPase systems.
Inexpressibility in Exp-Minus-Log
math.LO 2026-05 unverdicted novelty 6.0

In the Exp-Minus-Log system every expressible number is computable and Chaitin's Ω_U is inexpressible.
Algebraic structure behind Odrzywo{\l}ek's EML operator
math-ph 2026-04 unverdicted novelty 6.0

The EML operator forms an abelian group with functional inverses, giving a constructive algebraic route to many distinct families of transcendental elementary functions via binary trees.
Why Architecture Choice Matters in Symbolic Regression
cs.NE 2026-04 unverdicted novelty 6.0

Different fixed tree architectures in gradient-based symbolic regression produce dramatically different recovery rates, with more expressive structures sometimes failing where restricted ones succeed reliably.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages · cited by 5 Pith papers · 1 internal anchor

[1]

Press, Saul A

William H. Press, Saul A. Teukolsky, William T. Vetterling, and Brian P. Flannery. Numerical Recipes: The Art of Scientific Computing. Cambridge University Press, Cambridge, 3 edition, 2007. ISBN 978-0-521-88068-8

work page 2007
[2]

A set of five independent postulates for Boolean algebras, with application to logical constants.Transactions of the American Mathematical Society, 14(4):481–488, 1913

Henry Maurice Sheffer. A set of five independent postulates for Boolean algebras, with application to logical constants.Transactions of the American Mathematical Society, 14(4):481–488, 1913. doi: 10.1090/S0002-9947-1913-1500960-1

work page doi:10.1090/s0002-9947-1913-1500960-1 1913
[3]

Andrew Hart, Edinburgh, 1614

John Napier.Mirifici Logarithmorum Canonis Descriptio. Andrew Hart, Edinburgh, 1614

work page
[4]

London, 1624

Henry Briggs.Arithmetica Logarithmica. London, 1624

work page
[5]

Cambridge, 1722

Roger Cotes.Harmonia Mensurarum, sive Analysis & Synthesis per Rationum & An- gulorum Mensuras Promotae: Accedunt Alia Opuscula Mathematica. Cambridge, 1722. Posthumous; edited by Robert Smith

work page
[6]

M´ emoire sur l’int´ egration d’une classe de fonctions transcendantes

Joseph Liouville. M´ emoire sur l’int´ egration d’une classe de fonctions transcendantes. Journal f¨ ur die reine und angewandte Mathematik (Crelle’s Journal), 13:93–118, 1835. doi: 10.1515/crll.1835.13.93

work page doi:10.1515/crll.1835.13.93
[7]

Columbia University Press, New York, 1948

Joseph Fels Ritt.Integration in Finite Terms: Liouville’s Theory of Elementary Meth- ods. Columbia University Press, New York, 1948

work page 1948
[8]

De Gruyter Oldenbourg, Berlin, Boston, 2022

Bernd Ulmann.Analog Computing. De Gruyter Oldenbourg, Berlin, Boston, 2022. ISBN 9783110787740. doi: doi:10.1515/9783110787740. URLhttps://doi.org/10. 1515/9783110787740

work page doi:10.1515/9783110787740 2022
[9]

AI Feynman: A physics-inspired method for symbolic regression.Science Advances, 6(16):eaay2631, 2020

Silviu-Marian Udrescu and Max Tegmark. AI Feynman: A physics-inspired method for symbolic regression.Science Advances, 6(16):eaay2631, 2020. doi: 10.1126/sciadv. aay2631

work page doi:10.1126/sciadv 2020
[10]

Discovering symbolic models from deep learning with inductive biases.Advances in Neural Information Processing Systems, 33:17429–17442, 2020

Miles Cranmer, Alvaro Sanchez-Gonzalez, Peter Battaglia, Rui Xu, Kyle Cranmer, David Spergel, and Shirley Ho. Discovering symbolic models from deep learning with inductive biases.Advances in Neural Information Processing Systems, 33:17429–17442, 2020

work page 2020
[11]

Hou, and Max Tegmark

Ziming Liu, Yixuan Wang, Sachin Vaidya, Fabian Ruehle, James Halverson, Marin Soljaˇ ci´ c, Thomas Y. Hou, and Max Tegmark. KAN: Kolmogorov–Arnold networks. InInternational Conference on Learning Representations (ICLR), 2025. URLhttps: //openreview.net/forum?id=Ozo7qJ5vZi

work page 2025
[12]

Pergamon Press, Oxford, 1963

Jan Lukasiewicz.Elements of Mathematical Logic. Pergamon Press, Oxford, 1963. English translation ofElementy logiki matematycznej(1929)

work page 1963
[13]

H. S. Black. Stabilized feedback amplifiers.The Bell System Technical Journal, 13(1): 1–18, 1934. doi: 10.1002/j.1538-7305.1934.tb00652.x. 17

work page doi:10.1002/j.1538-7305.1934.tb00652.x 1934
[14]

Vinod Nair and Geoffrey E. Hinton. Rectified linear units improve restricted boltzmann machines. InProceedings of the 27th International Conference on Machine Learning (ICML 2010), pages 807–814, 2010

work page 2010
[15]

MIT Press, Cambridge, MA, 2016

Ian Goodfellow, Yoshua Bengio, and Aaron Courville.Deep Learning. MIT Press, Cambridge, MA, 2016. ISBN 9780262035613. URLhttps://www.deeplearningbook. org/

work page 2016
[16]

Wolfram Media, 2002

Stephen Wolfram.A New Kind of Science. Wolfram Media, 2002. ISBN 1-57955-008-8. URLhttps://www.wolframscience.com/nks/

work page 2002
[17]

Sch¨ onfinkel.¨Uber die bausteine der mathematischen logik.Mathematische Annalen, 92(3):305–316, 1924

M. Sch¨ onfinkel.¨Uber die bausteine der mathematischen logik.Mathematische Annalen, 92(3):305–316, 1924. ISSN 1432-1807. doi: 10.1007/BF01448013. URLhttps://doi. org/10.1007/BF01448013

work page doi:10.1007/bf01448013 1924
[18]

Curry and R

H.B. Curry and R. Feys.Combinatory Logic. Number v. 1 in Combinatory Logic. North-Holland Publishing Company, 1958. URLhttps://books.google.pl/books? id=fEnuAAAAMAAJ

work page 1958
[19]

Interaction combinators.Inf

Yves Lafont. Interaction combinators.Inf. Comput., 137(1):69–101, August 1997. ISSN 0890-5401. doi: 10.1006/inco.1997.2643. URLhttps://doi.org/10.1006/inco.1997. 2643

work page doi:10.1006/inco.1997.2643 1997
[20]

On the sheffer stroke operation in fuzzy logic.Fuzzy Sets and Systems, 431:110–128, 2022

Micha l Baczy´ nski, Pedro Berruezo, Piotr Helbin, Sebastia Massanet, Wanda Niemyska, and Daniel Ruiz-Aguilera. On the sheffer stroke operation in fuzzy logic.Fuzzy Sets and Systems, 431:110–128, 2022. ISSN 0165-0114. doi: https://doi.org/10. 1016/j.fss.2021.05.003. URLhttps://www.sciencedirect.com/science/article/ pii/S0165011421001834. Logic and Related Topics

work page 2022
[21]

A simple multi-processor computer based on subleq,

Oleg Mazonka and Alex Kolodin. A simple multi-processor computer based on subleq,

work page
[22]

URLhttps://arxiv.org/abs/1106.2593

work page internal anchor Pith review Pith/arXiv arXiv
[23]

J. H. Conway.FRACTRAN: A Simple Universal Programming Language for Arithmetic, pages 4–26. Springer New York, New York, NY, 1987. ISBN 978-1-4612-4808-8. doi: 10.1007/978-1-4612-4808-8 2. URLhttps://doi.org/10.1007/978-1-4612-4808-8_ 2

work page doi:10.1007/978-1-4612-4808-8 1987
[24]

Universality in elementary cellular automata.Complex Systems, 15(1): 1–40, 2004

Matthew Cook. Universality in elementary cellular automata.Complex Systems, 15(1): 1–40, 2004. doi: 10.25088/ComplexSystems.15.1.1

work page doi:10.25088/complexsystems.15.1.1 2004
[25]

Linsky.The Evolution of Principia Mathematica: Bertrand Russell’s Manuscripts and Notes for the Second Edition

B. Linsky.The Evolution of Principia Mathematica: Bertrand Russell’s Manuscripts and Notes for the Second Edition. Cambridge University Press, 2011. ISBN 9781139497336. URLhttps://books.google.pl/books?id=EHCMuK4gS-MC

work page 2011
[26]

Francis H. C. Crick and James D. Watson. Genetical implications of the structure of deoxyribonucleic acid.Nature, 171(4361):964–967, 1953. doi: 10.1038/171964b0. 18

work page doi:10.1038/171964b0 1953
[27]

Smith, J

D. Smith, J. S. Myers, C. S. Kaplan, and C. Goodman-Strauss. An aperiodic monotile.Combinatorial Theory, 4(1), 2024. doi: 10.5070/C64163843. URLhttps: //escholarship.org/uc/item/3317z9z9

work page doi:10.5070/c64163843 2024
[28]

How to build the perfect igloo.Eureka, 63:1001110–1001111, 2014

Andrzej Odrzywolek. How to build the perfect igloo.Eureka, 63:1001110–1001111, 2014

work page 2014
[29]

Fehlberg

E. Fehlberg. Klassische runge-kutta-formeln vierter und niedrigerer ordnung mit schrittweiten-kontrolle und ihre anwendung auf w¨ armeleitungsprobleme.Comput- ing, 6(1):61–71, 1970. ISSN 1436-5057. doi: 10.1007/BF02241732. URLhttps: //doi.org/10.1007/BF02241732

work page doi:10.1007/bf02241732 1970
[30]

Double exponential formulas for numerical inte- gration.Publications of the Research Institute for Mathematical Sciences, 9(3):721–741,

Hidetosi Takahasi and Masatake Mori. Double exponential formulas for numerical inte- gration.Publications of the Research Institute for Mathematical Sciences, 9(3):721–741,

work page
[31]

doi: 10.2977/prims/1195192451

work page doi:10.2977/prims/1195192451
[32]

Cooley and John W

James W. Cooley and John W. Tukey. An algorithm for the machine calculation of complex Fourier series.Mathematics of Computation, 19(90):297–301, 1965. doi: 10. 1090/S0025-5718-1965-0178586-1

work page 1965
[33]

Growth estimates for exp—log functions.Journal of Symbolic Computation, 10(6):611–632, 1990

John Shackell. Growth estimates for exp—log functions.Journal of Symbolic Computation, 10(6):611–632, 1990. ISSN 0747-7171. doi: https://doi.org/10.1016/ S0747-7171(08)80161-7. URLhttps://www.sciencedirect.com/science/article/ pii/S0747717108801617

work page 1990
[34]

Asymptotic expansions of exp-log functions

Daniel Richardson, Bruno Salvy, John Shackell, and Joris van der Hoeven. Asymptotic expansions of exp-log functions. Technical Report RR-2859, INRIA, 1996. URLhttps: //inria.hal.science/inria-00073832. inria-00073832

work page 1996
[35]

Mathematica, Version 14.3, 2025

Wolfram Research, Inc. Mathematica, Version 14.3, 2025. URLhttps://www.wolfram. com/mathematica. Champaign, IL, 2025

work page 2025
[36]

Broken calculator warm up.https://mathforlove.com/lesson/ broken-calculator-warmup/, 2017

Dan Finkel. Broken calculator warm up.https://mathforlove.com/lesson/ broken-calculator-warmup/, 2017. Accessed 2025-11-18

work page 2017
[37]

Newman.A Problem Seminar

Donald J. Newman.A Problem Seminar. Sources in the History of Mathematics and Physical Sciences. Springer-Verlag, New York, 1982. ISBN 0-387-90765-3

work page 1982
[38]

Mathematical discovery in the age of artifi- cial intelligence.Nature Physics, 21(10):1504–1506, October 2025

Bartosz Naskrecki and Ken Ono. Mathematical discovery in the age of artifi- cial intelligence.Nature Physics, 21(10):1504–1506, October 2025. doi: 10.1038/ s41567-025-03042-0. URLhttps://doi.org/10.1038/s41567-025-03042-0

work page doi:10.1038/s41567-025-03042-0 2025
[39]

A. N. Kolmogorov. Three approaches to the quantitative definition of information. Problems of Information Transmission, 1(1):1–7, 1965

work page 1965
[40]

Some consequences of schanuel’s conjecture in exponential rings

Giuseppina Terzo. Some consequences of schanuel’s conjecture in exponential rings. Communications in Algebra, 36(3):1171–1189, 2008. doi: 10.1080/00927870701410694. URLhttps://doi.org/10.1080/00927870701410694. 19

work page doi:10.1080/00927870701410694 2008
[41]

SymbolicRegressionPackage: Basic building blocks for symbolic regression.https://github.com/VA00/SymbolicRegressionPackage, 2026

Andrzej Odrzywo lek. SymbolicRegressionPackage: Basic building blocks for symbolic regression.https://github.com/VA00/SymbolicRegressionPackage, 2026. Includes EML toolkit

work page 2026
[42]

R., Millman, K

Charles R. Harris, K. Jarrod Millman, St´ efan J. van der Walt, Ralf Gommers, Pauli Virtanen, David Cournapeau, Eric Wieser, Julian Taylor, Sebastian Berg, Nathaniel J. Smith, Robert Kern, Matti Picus, Stephan Hoyer, Marten H. van Kerkwijk, Matthew Brett, Allan Haldane, Jaime Fern´ andez del R´ ıo, Mark Wiebe, Pearu Peterson, Pierre G´ erard-Marchant, Kev...

work page doi:10.1038/s41586-020-2649-2 2020
[43]

Py- Torch: An imperative style, high-performance deep learning library

Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmai- son, Andreas K¨ opf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu Fang, Junjie Bai, and Soumith Chintala. Py- Torch: An imperative style, high...

work page 2019
[44]

The lean 4 theorem prover and programming language

Leonardo de Moura and Sebastian Ullrich. The lean 4 theorem prover and programming language. InAutomated Deduction – CADE 28: 28th International Conference on Automated Deduction, Virtual Event, July 12–15, 2021, Proceedings, pages 625–635, Berlin, Heidelberg, 2021. Springer-Verlag. doi: 10.1007/978-3-030-79876-5 37. URL https://doi.org/10.1007/978-3-030-...

work page doi:10.1007/978-3-030-79876-5 2021
[45]

Beyond zeros and ones – analog computing in the twenty-first cen- tury.International Journal of Parallel, Emergent and Distributed Systems, 39(2):139– 151, 2024

Bernd Ulmann. Beyond zeros and ones – analog computing in the twenty-first cen- tury.International Journal of Parallel, Emergent and Distributed Systems, 39(2):139– 151, 2024. doi: 10.1080/17445760.2023.2296672. URLhttps://doi.org/10.1080/ 17445760.2023.2296672

work page doi:10.1080/17445760.2023.2296672 2024
[46]

Interpretable machine learning for science with PySR and SymbolicRe- gression.jl, 2023

Miles Cranmer. Interpretable machine learning for science with PySR and SymbolicRe- gression.jl, 2023

work page 2023
[47]

Categorical reparameterization with Gumbel- Softmax

Eric Jang, Shixiang Gu, and Ben Poole. Categorical reparameterization with Gumbel- Softmax. InProceedings of the 5th International Conference on Learning Representa- tions (ICLR), 2017. URLhttps://openreview.net/forum?id=rkE3y85ee

work page 2017
[48]

Petersen, Mikel Landajuela, T

Brenden K. Petersen, Mikel Landajuela, T. Nathan Mundhenk, Cl´ audio P. Santiago, Sookyung Kim, and Joanne T. Kim. Deep symbolic regression: Recovering mathemati- cal expressions from data via risk-seeking policy gradients. InInternational Conference on Learning Representations (ICLR), 2021. URLhttps://openreview.net/forum? id=m5Qsh0kBQG

work page 2021
[49]

A ternary Sheffer operator for elementary functions?Acta Physica Polonica B, 2026

Andrzej Odrzywolek. A ternary Sheffer operator for elementary functions?Acta Physica Polonica B, 2026. in preparation. 20

work page 2026
[50]

Activation func- tions in deep learning: A comprehensive survey and benchmark.Neurocomputing, 503: 92–108, 2022

Shiv Ram Dubey, Satish Kumar Singh, and Bidyut Baran Chaudhuri. Activation func- tions in deep learning: A comprehensive survey and benchmark.Neurocomputing, 503: 92–108, 2022. doi: 10.1016/j.neucom.2022.06.111

work page doi:10.1016/j.neucom.2022.06.111 2022
[51]

Deep learning for symbolic mathematics

Guillaume Lample and Fran¸ cois Charton. Deep learning for symbolic mathematics. In Proceedings of the 8th International Conference on Learning Representations (ICLR),

work page
[52]

URLhttps://openreview.net/forum?id=S1eZYeHFDS

work page
[53]

phylogenetic

Nayoung Lee, Kartik Sreenivasan, Jason D. Lee, Kangwook Lee, and Dimitris Papail- iopoulos. Teaching arithmetic to small transformers. InThe Twelfth International Conference on Learning Representations (ICLR), 2024. URLhttps://openreview. net/forum?id=dsUB4bst9S. 21 eml 1e exp ln − −1 2 −x + 1/x × x2 ÷x/2 x+y 2 √x xy logxy π √x2+y2 σ cosh sinh tanh cos si...

work page 2024