M.J. Weinberger, J. Rissanen, et al.
ACSSC 1995
Consider a vector-valued stationary random process {Yk}∞-∞, from which the estimates, R0, R1, ..., RN, of the covariance matrices EYkY′k-i, i=0, 1, 1, ..., N, can be made. The three matrices, (G(N), F(N), H(N)) are said to form a minimal partial realization of the process {Yk}∞-∞ if we can write. {A figure is presented}. {A figure is presented}. In the paper an algorithm is described for recursively calculating the minimal partial repreentations for the sequences R0, ..., RN, N=0, 1, ..., taken as inputs to the algorithm.
M.J. Weinberger, J. Rissanen, et al.
ACSSC 1995
J. Rissanen
ISIT 2000
J. Rissanen, G.S. Shedler
Journal of Statistical Planning and Inference
J. Rissanen
IEEE Trans. Inf. Theory