Word
markovian
Phoneticsmar·kov·ian
/ˌmɑɹˈkoʊˌviən/
Definition
adj. describing a process or system where the next state depends only on where it is right now, not on its past history.
adj. possessing the Markov property, whereby the conditional probability distribution of future states depends solely upon the present state and not on the preceding sequence of events.
Examples
“A game of board dice is Markovian because your next turn depends only on your current square, not how you got there.”
“Reinforcement learning algorithms typically assume a Markovian decision process where the agent's reward is conditioned solely on the immediate state and action.”
“Non-Markovian dynamics introduce memory kernels into the master equation, significantly complicating the convergence guarantees of standard stochastic optimization algorithms.”
Examples
simple
“A game of board dice is Markovian because your next turn depends only on your current square, not how you got there.”
contextual
“Reinforcement learning algorithms typically assume a Markovian decision process where the agent's reward is conditioned solely on the immediate state and action.”
complex
“Non-Markovian dynamics introduce memory kernels into the master equation, significantly complicating the convergence guarantees of standard stochastic optimization algorithms.”
Etymology
Named after Russian mathematician Andrey Andreyevich Markov (1856–1922), who introduced Markov chains in 1906. Markov famously verified his mathematical theory by manually analyzing a chain of 20,000 consonants and vowels in Alexander Pushkin's verse novel Eugene Onegin to show that the law of large numbers could apply to mutually dependent events.
Etymology adapted from Wiktionary, available under CC BY-SA 4.0.