Markov Chain

Linear
Graphical Model without
Collider, Markovian

A sequence with a memoryless property with several chained states

Markov chains are abstractions of

Non-Deterministic Turing Machine). A Markov chain consists of states, plus an transition probability matrix .

At each step, we are in exactly one of the states

For , the matrix entry tells us the probability of being the next state, given we are currently in state where .

Markov chain is a model of stochastic evolution of the system captured in discrete snapshots. The

Stochastic matrix describes the probabilities with which the system transits into different state. Therefore, You can start with a messy process which is not

Stationary process but which will eventually converge to a well behaved

Stationary process which is driven by only one probability law and your process can freely visit all states (

Ergodicity) within state spaces without getting trapped in a loop.

The point is that the probability of a state at a particular time depends only on the immediately preceding state, following the Markov assumption, which is a discrete-time stochastic process.

At any point of the sequence, the marginal distribution is given by

Markov Chain Notion

Transition probability matrix

Second-order Markov chain

Markov Reward Model

Irreducibility

Aperiodicity

Time-homogeneous Markov

Markov Chain Ergodicity

[Machine learning] Markov Chain, Gibbs Sampling, 마르코프 체인, 깁스 샘플링 (day2 / 201010)

Q. Markov Chain을 고등학생에게 설명한다면 어떤 방식이 좋을까요? Q. Markov Chain 은 머신러닝 알고리즘 중 어디에 활용이 되나요? Q. 깁스 샘플링 은 무엇인가 Q. 깁스 샘플링 은 왜 쓰는가 ? sites.google.com/site/machlearnwiki/RBM/markov-chain (제가 보려고 여기에 다시 옮겨 적어유 원문은 링크로 ! ) 마코프 체인은 마코프 성질을 지닌 이산 확률 과정이라고 합니다.