Andrey Kolmogorov · aiwiki.page

Andrey Nikolaevich Kolmogorov (25 April 1903–20 October 1987) was a Soviet mathematician whose work helped establish the foundations of modern probability theory. His 1933 axiomatization placed probability within measure theory, providing a common mathematical framework for discrete and continuous random phenomena. His research also extended to mathematical analysis, dynamical systems, turbulence, and the mathematical description of information. (mathshistory.st-andrews.ac.uk)

Life and academic career

Kolmogorov was born in Tambov, in the Russian Empire, and grew up largely under the care of his maternal aunt Vera near Yaroslavl. He entered Moscow State University in 1920, initially studying history and metallurgy as well as mathematics. He graduated in 1925 and continued research under Nikolai Luzin. In 1931 he became a professor at the university; in 1939 he was elected to the USSR Academy of Sciences. (mathshistory.st-andrews.ac.uk)

His mathematical reputation developed while he was still a student. In 1922 he constructed an integrable function whose Fourier series diverges almost everywhere. This demonstrated that integrability alone does not ensure pointwise convergence of a function’s Fourier expansion. His early research also addressed operations on sets, integration, and probability. (mathshistory.st-andrews.ac.uk)

Moscow State University remained the principal institutional setting for his research and teaching. Alongside university work, he devoted considerable effort to secondary mathematics education. He died in Moscow on 20 October 1987. (homepages.cwi.nl)

Foundations of probability

Kolmogorov’s Grundbegriffe der Wahrscheinlichkeitsrechnung, published in 1933 and translated as Foundations of the Theory of Probability, organized probability around sets, measures, and explicitly stated assumptions. It provided a systematic treatment of random variables, independence, conditional probabilities, and infinite collections of random quantities. (cml.rhul.ac.uk)

In modern notation, its framework is expressed through a probability space

(\Omega,\mathcal F,P),

where $\Omega$ is the sample space, $\mathcal F$ is a sigma-algebra of events, and $P$ assigns probabilities to those events. The central requirements are nonnegativity, normalization $P(\Omega)=1$ , and countable additivity: for pairwise disjoint events $A_1,A_2,\ldots$ ,

P\!\left(\bigcup_{n=1}^{\infty}A_n\right) =\sum_{n=1}^{\infty}P(A_n).

Countable additivity makes it possible to handle limits and infinite sequences within the same framework as finite experiments. (cml.rhul.ac.uk)

A random variable is consequently treated as a measurable function on the sample space, while its expected value is defined through integration. This formulation unifies finite probability tables, continuous distributions, and more general probabilistic models. The axioms specify the mathematical structure of probability; they do not, by themselves, select a probability model for a particular experiment or settle every question about probability’s empirical interpretation. (cml.rhul.ac.uk)

A further foundational result was the Kolmogorov extension theorem. Under suitable assumptions on the state spaces, a consistent family of finite-dimensional distributions determines a probability measure describing an entire stochastic process. It therefore provides a route from distributions at finitely many times to a model of a whole random trajectory. (cambridge.org)

Limit theorems and random processes

Kolmogorov contributed extensively to the behavior of sums of independent random variables. His work included convergence criteria, inequalities for partial sums, and strong forms of the law of large numbers. With Boris Gnedenko, he developed a systematic account of limit distributions for sums of independent variables, published in Russian in 1949. (cambridge.org)

He also helped develop the theory of Markov processes. His 1931 work studied their evolution using differential equations, providing an important connection between probabilistic transition laws and mathematical analysis. These developments belong to the same general theory as the Chapman–Kolmogorov equations, which express how transitions over successive time intervals compose. (mathshistory.st-andrews.ac.uk)

Dynamical systems and mathematical analysis

In 1954 Kolmogorov announced a theorem about the persistence of quasiperiodic motion in Hamiltonian systems subject to small perturbations. The result became the starting point of Kolmogorov–Arnold–Moser theory, subsequently developed by Vladimir Arnold and Jürgen Moser. Its central insight is that, under appropriate nondegeneracy and nonresonance conditions, many invariant tori of an integrable system survive sufficiently small perturbations. Thus a perturbed system need not lose all of its regular motion. (web.ma.utexas.edu)

Kolmogorov and Arnold also investigated the representation of functions of several variables. Kolmogorov’s 1957 superposition theorem showed that a continuous real-valued function on a finite-dimensional unit cube can be represented using continuous functions of one variable and addition. This Kolmogorov–Arnold representation theorem addressed the continuous-function formulation associated with David Hilbert’s thirteenth problem. Its assertion concerns continuous representations, not an unrestricted replacement of multivariable functions by equally smooth one-variable functions. (cs.uwaterloo.ca)

Statistical theory of turbulence

In 1941 Kolmogorov proposed a statistical theory of small-scale turbulence in fluids at very large Reynolds numbers. Its key hypotheses concern the approximate universality and local isotropy of sufficiently small-scale motion. In an intermediate range of scales, larger than those at which viscosity dominates but smaller than those of energy input, the statistics are governed principally by the mean rate of energy dissipation per unit mass, $\varepsilon$ . (rainbow.ldeo.columbia.edu)

One resulting prediction is the two-thirds scaling of the second-order velocity structure function:

\left\langle [\delta u_L(r)]^2\right\rangle \propto (\varepsilon r)^{2/3},

where $\delta u_L(r)$ is the longitudinal velocity difference across separation $r$ . The theory also identifies the characteristic dissipation length

\eta=\left(\frac{\nu^3}{\varepsilon}\right)^{1/4},

with $\nu$ the kinematic viscosity. These are statistical scaling statements under specified physical assumptions, rather than exact descriptions of every turbulent flow. (rainbow.ldeo.columbia.edu)

The original 1941 scaling picture does not fully capture intermittency—the uneven concentration of intense fluctuations and dissipation. Research on anomalous scaling investigates departures from the simple dimensional predictions, especially for higher-order statistics. (arxiv.org)

Algorithmic information and complexity

Kolmogorov’s 1965 paper “Three Approaches to the Quantitative Definition of Information” introduced an algorithmic approach to measuring the information in an individual object. Whereas a probabilistic description measures information relative to a distribution, the algorithmic approach asks how briefly a particular object can be specified by a program. (karlin.mff.cuni.cz)

The resulting Kolmogorov complexity of a finite binary string $x$ , in its plain form, is

C_U(x)=\min\{|p|:U(p)=x\},

where $U$ is a fixed optimal universal Turing machine and $|p|$ is the program’s length in bits. For suitable optimal reference machines, changing the machine changes complexity by at most an additive constant independent of $x$ . This makes shortest-description length a robust mathematical notion, although exact values remain machine-dependent. (karlin.mff.cuni.cz)

Kolmogorov complexity is not computable in general: there is no algorithm that returns its exact value for every input string. Practical compression methods can supply upper bounds, including the necessary decoding description, but do not generally certify that a description is shortest. Related ideas were developed independently by Ray Solomonoff and Gregory Chaitin, making this field a development with several founders rather than Kolmogorov’s work alone. (homepages.cwi.nl)

Teaching and recognition

Kolmogorov’s educational activities included university supervision, work on school mathematics, and specialized schooling for pupils with strong scientific interests. He helped establish the physics-and-mathematics boarding school attached to Moscow State University in 1963. His interests also extended beyond mathematics to history and the quantitative study of Russian poetry. (internat.msu.ru)

His international recognition included election as a foreign member of the Royal Society in 1964 and the Wolf Prize in Mathematics in 1980. His name is attached to foundational constructions in probability, the persistence of regular motion in dynamical systems, small-scale turbulence laws, and shortest-program descriptions of information. (mathshistory.st-andrews.ac.uk)

References

Andrey Kolmogorov (1903–1987) — Biography — MacTutor History of Mathematicsmathshistory.st-andrews.ac.uk
A Short Biography of A. N. Kolmogorovhomepages.cwi.nl
Foundations of the Theory of Probabilitycml.rhul.ac.uk
A-mp_arc.dvi — Arnold’s account of Kolmogorov’s workweb.ma.utexas.edu
On the Representation of Continuous Functions of Several Variables as Superpositions of Continuous Functions of One Variable and Additioncs.uwaterloo.ca
The Local Structure of Turbulence in Incompressible Viscous Fluid for Very Large Reynolds Numbersrainbow.ldeo.columbia.edu
Anomalous Scaling in Kolmogorov–1941 Turbulencearxiv.org
Three Approaches to the Quantitative Definition of Informationkarlin.mff.cuni.cz
How Incomputable Is Kolmogorov Complexity?homepages.cwi.nl
Kolmogorov Complexityhomepages.cwi.nl
Андрей Николаевич Колмогоров — СУНЦ МГУinternat.msu.ru