The story
In 1906 the Russian mathematician Andrey Markov, working in St. Petersburg, published a paper extending the law of large numbers to quantities that depend on one another. He studied sequences in which the next state depends only on the present one, not on how the sequence got there. These are now called Markov chains. A simple example is weather where tomorrow's chance of rain depends only on whether it rains today.
Probability theory had until then mostly treated events as independent, like coin flips. A rival, Pavel Nekrasov, argued that the law of large numbers requires independence and used that claim in an argument about free will. Markov, a sharp-tongued critic, showed by mathematical proof that averages still settle down for dependent chains. In 1913 he applied the idea to the first 20,000 letters of Pushkin's Eugene Onegin, counting how often a vowel followed a vowel or a consonant. The mathematics proved far more important than the quarrel. Markov chains power models of language and speech in hidden Markov models, Google's PageRank algorithm, Monte Carlo simulation methods used in physics and statistics, and the planning methods of modern reinforcement learning.
Why it mattered
- Hidden Markov models drove early speech recognition and gene-finding software.
- Google's PageRank treats a web surfer as a Markov chain, ranking pages by where the surfer ends up.
- Markov chain Monte Carlo methods became a core tool of statistics and physics.
- Claude Shannon's models of English text built on the same idea and fed information theory.
Sources
- Andrey Andreyevich Markov Britannica
- First Links in the Markov Chain American Scientist
- Andrei Andreyevich Markov (1856-1922) MacTutor, University of St Andrews
Drafted by AI (Claude) from the sources above and checked against them by AI; no historian has reviewed this page. Spot a mistake? Suggest a correction. How pages are made
Community notes
Loading notes…