On minimal representations of rational arrival processes (Q1945559): Difference between revisions

Let \(\nu \) be a row vector of size \(m\) and \({D_0}\) be a \(m \times m\) matrix. Then \((\nu ,{D_0})\) defines a matrix exponential distribution, if and only if \({F_{(\nu ,{D_0})}}(t) = 1 - \nu {e^{{D_0}t}}{\mathbf{1}}\) is a valid distribution function where \({\mathbf{1}}\) on the right-hand side is the unit column vector of appropriate size. Different representations may exist that describe the same distribution. Two representations \((\nu,{D_0})\) and \((\phi,{C_0})\) are said to be equivalent if and only if \({F_{(\nu ,{D_0})}}(t) \equiv {F_{(\phi,{C_0})}}(t)\). A rational arrival process (RAP) is a point process whose joint density of the consecutive interarrival times is a matrix exponential function of the variables and the double transform of the number of arrivals is a rational function of the transform variables. The paper presents an approach to determine the equivalence of different representations of RAPs and gives necessary and sufficient conditions for the size of minimal representations for RAPs.