en.wikipedia.org/wiki/Kolmogorov_complexity
2 corrections found
It is named after Andrey Kolmogorov, who first published on the subject in 1963
Kolmogorov is the namesake, but the standard historical sources place his introduction of the complexity concept in 1964 and his first dedicated publication in 1965, not 1963.
Full reasoning
The date here is off.
A standard survey by Zvonkin and Levin states: "In 1964 Kolmogorov introduced the concept of the complexity of a finite object". That same survey treats Kolmogorov's complexity work as a mid-1960s development, not a 1963 publication.
Math-Net's bibliographic record for Kolmogorov's foundational paper "Three approaches to the definition of the concept 'quantity of information'" shows it was published in 1965. By contrast, the 1963 paper listed in the references is "On tables of random numbers", which is a different publication.
So the article is right that the notion is named after Kolmogorov, but "first published on the subject in 1963" is not consistent with the cited historical record; the key dates are 1964 for introducing the concept and 1965 for the first dedicated publication.
3 sources
- A. K. Zvonkin, L. A. Levin, "The complexity of finite objects and the development of the concepts of information and randomness by means of the theory of algorithms"
Abstract: In 1964 Kolmogorov introduced the concept of the complexity of a finite object ...
- A. N. Kolmogorov, "Three approaches to the definition of the concept \"quantity of information\""
Problemy Peredachi Informatsii, 1965, Volume 1, Issue 1, Pages 3-11 ... Received: 09.01.1965
- On tables of random numbers
Kolmogorov, A. N. Sankhya, the Indian Journal of Statistics: Series A, 25(4):369-376, 1963.
K(X,Y) = K(X) + K(Y|X) + c*max(1,log(K(X,Y))).
The chain rule is not an exact equality with a fixed added positive term; it holds only up to an additive logarithmic error, usually written as ±O(log K(X,Y)) or O(log n).
Full reasoning
As written, this formula is mathematically wrong.
The standard chain rule / symmetry-of-information statement is that the joint complexity equals the sum of the marginal and conditional complexities up to an additive logarithmic error. Modern statements write this as
- K(x,y) = K(x) + K(y|x) ± O(log K(x,y)), or equivalently
- K(x,y) ≤ K(x) + K(y|x) + O(log n) and K(x,y) ≥ K(x) + K(y|x) - O(log n).
That is very different from an exact equality with a specific positive term + c*max(1, log(K(X,Y))). The latter fixes the sign and size of the error instead of allowing a bounded additive discrepancy. So the article should use an O(log ...) / ± O(log ...) formulation, not the displayed exact equality.
3 sources
- ECCC TR26-021: Failure of Symmetry of Information for Randomized Computations
Symmetry of Information ... states that for all n-bit strings x and y, K(x,y) = K(y) + K(x|y) up to an additive error of O(log n).
- Electronic Colloquium on Computational Complexity, Revision 1 of Report No. 215 (2025)
K(x, y) ≤ K(x) + K(y | x) + O(log n), ... K(x, y) ≥ K(x) + K(y | x) - O(log n).
- Kolmogorov Complexity Characterizes Statistical Zero Knowledge (ITCS 2023)
Theorem 2 (Symmetry of Information). K(x, y) = K(x) + K(y|x) ± O(log(K(x, y))).