We develop a theory of minimal realizations of a finite sequence over
an integral domain R, from first principles. Our notion of a minimal r
ealization is closely related to that of a linear recurring sequence a
nd of a partial realization (as in Mathematical Systems Theory). From
this theory, we derive Algorithm MR which computes a minimal realizati
on of a sequence of L elements using at most L(5L + 1)/2 R-multiplicat
ions, Wa also characterize all minimal realizations of a given sequenc
e in terms of the computed minimal realization. This algorithm compute
s the linear complexity of an R sequence, solves non-singular linear s
ystems over R (extending Wiedemann's method), computes the minimal pol
ynomial of an R-matrix, transfer/growth functions and symbolic Pade ap
proximations. There are also a number of applications to Coding Theory
. We thus provide a common framework for solving some well-known probl
ems in Systems Theory, Symbolic/Algebraic Computation and Coding Theor
y.