CompetingPoisson(rates; events)

The first arrival from competing Poisson point processes with the given rates.

rand returns a NamedTuple{(:duration,:event)}, where duration is the duration until the first event and event is drawn from events.

Varios properties are supported, see propertynames.

Note

The distribution of duration is Exponential(1/sum(rates)), so this process can also be thought of as competing exponentials.

TransitionRateMatrix(Q)

Create a transition rate matrix from the argument. This makes a copy, checks that off-diagonal elements are nonnegative, and sets the diagonal so that rows sum to 0, striving to preserve sparsity and structure.

TransitionRateMatrix!(Q)

Create a TransitionRateMatrix from a matrix Q. Modifies the argument (to normalize the diagonal).

_off_diagonal_checked_row_sum!(d, A)

Accumulate the sum of off-diagonal elements into d by row. Return d.

Internal, not part of the API. Works with generalized indexing.

checked_square_axis(A)

Assert that both axes are the same, and return a single one.

Internal, not part of the API. Works with generalized indexing.

normalize_diagonal!(A)

Set the diagonal of the argument (which is modified) so that rows sum to 0.

Internal, not part of the API. Works with generalized indexing.

off_diagonal_checked_row_sum(A)

Check that off-diagonal elements are nonnegative, and return their sum by row.

Internal, not part of the API. Works with generalized indexing.

stationary_distribution(m)

Return the stationary distribution of a transition rate matrix as a vector.