aeppl.dists.discrete_markov_chain(Gammas, gamma_0, size=None, srng=None, **kwargs)[source]#

Construct a first-order discrete Markov chain distribution.

This characterizes vector random variables consisting of state indicator values (i.e. 0 to M - 1) that are driven by a discrete Markov chain.

  • Gammas (TensorVariable) – An array of transition probability matrices. Gammas takes the shape ... x N x M x M for a state sequence of length N having M-many distinct states. Each row, r, in a transition probability matrix gives the probability of transitioning from state r to each other state.

  • gamma_0 (TensorVariable) – The initial state probabilities. The last dimension should be length M, i.e. the number of distinct states.