Saddle Point
Back to the cover

Part I · Foundations

  1. 0 Mathematical Foundations
  2. 1 Observation, Beliefs, and Decision

Part II · Engines

  1. 2 Competing for Allocation: The Blotto Game
  2. 3 Bargaining, Commitment, and Repeated Cooperation
  3. 4 Signals, Disclosure, and Persuasion
  4. 5 Different Models of One Game
  5. 6 Leadership and Rules: Inspection
  6. 7 Decisions in Time
  7. 8 Many Agents
  8. 9 Networks, Dependence, and Resilience
  9. 10 Aggregating Information

Part III · Systems

  1. 11 Architecture of Decision Support
  2. 12 Simulation and Wargaming with AI Agents

Part IV · Limits

  1. 13 Where the Mathematics Stops

Chapter 8 · Many Agents

Mikhail L. Tsetlin

1924 – 1966

The problem on the desk

Can a finite device learn useful behaviour from rewards or penalties in an unknown random environment? Tsetlin analysed how internal states and transition rules affect long-run expected penalty.1

A life in moments

  1. 1924

    Moscow

    Born on 22 September.2

  2. 1941

    Moscow State University · Moscow

    Entered the physics faculty.2

  3. 1952

    Moscow State University · Moscow

    Completed his studies in the theory of oscillations.2

  4. 1958

    Moscow

    Defended his candidate dissertation.2

  5. 1961

    Published the finite-automata paper on random media.3

  6. 1964

    Moscow

    Defended his doctoral dissertation in May.2

  7. 1966

    Died on 30 May.2

Every line rests on a dated source (numbered below) unless it carries a mark: R for an inference, S for a story as told.

In their own words

The notion of purposeful behaviour of an automaton is introduced.

Вводится понятие целесообразности поведения автомата

On the Behaviour of Finite Automata in Random Media, 1961; translation ours5

A story

When more memory can hurt

Tsetlin contrasted a stationary environment with a changing Markov environment. In the latter, an automaton may need a finite optimal memory: increasing memory is not automatically an improvement.4

Around Tsetlin

  • co-author Israel Gelfand Gelfand and Tsetlin co-authored Some Methods of Control of Complex Systems in 1962.6

The genealogy lists only what the database records; a missing name is not a missing teacher.

The card in the book

Mathematician and biophysicist, a Moscow State University graduate of 1952, who worked in Moscow, from 1958 in Israel Gelfand's seminar on the boundary of mathematics and physiology. His 1961 paper pioneered the study of finite automata in random media — how trivially simple devices can adapt to unknown environments — and the games of collectives of such automata, including the Goore game proposed in his circle in 1963 and published by others in 1965, showed that collectives can concentrate on an optimum of a function none of them observes; the line was continued by Varshavsky's work on the collective behaviour of automata. His papers were collected posthumously (Nauka, 1969; English edition 1973).

Main works “On the behaviour of finite automata in random media” (Avtomatika i telemekhanika 22(10), 1961, pp. 1345–1354); Automaton Theory and Modeling of Biological Systems (Academic Press, 1973; Russian original, Nauka, 1969).

To read further

Sources (6)
  1. On the Behaviour of Finite Automata in Random Media, M. L. Tsetlin, 1961; pp. 1345–1347
  2. Mikhail Lvovich Tsetlin (obituary), 1966; p. 183
  3. On the Behaviour of Finite Automata in Random Media, M. L. Tsetlin, 1961; p. 1345
  4. On the Behaviour of Finite Automata in Random Media, M. L. Tsetlin, 1961; Abstract and introduction, pp. 1345–1346
  5. On the Behaviour of Finite Automata in Random Media, M. L. Tsetlin, 1961; p. 1345, abstract
  6. Mikhail Lvovich Tsetlin: publications, Math-Net.Ru; 1962 entry: I. M. Gelfand and M. L. Tsetlin, Some methods of control of complex systems

The facts on this page were gathered for the site from the sources listed, each checked against its locator; the book's card is reproduced unchanged.