The solution y(x) is unique only if the initial conditions are set. Let y(0) = c0, y′(0) = c1, y′′(0) = c2, and so on up to n-1. Given these initial constants, a deterministic process builds y(1), y(2), y(3), and so on.
Since the differences at 0 are known, up to order n-1, plug them into f and find the value of the nth difference at 0. Call this cn.
If w is the n-1st difference sequence evaluated at 1, write cn = w-cn-1 and solve for w. Do this all the way down the line, until y(1) = y′(0)+y(0).
The difference sequences up to order n-1 have been evaluated at x = 1; apply f and find the nth difference, evaluated at 1. This allows us to compute the lesser differences at x = 2, then the nth difference at x = 2, then the lesser differences at x = 3, and so on, building the unique solution y(x).
There are other ways to specify the initial conditions. For instance, specifying y(0) through y(n-1) provides enough information to construct the difference values at 0, up through order n-1. This ensures a unique solution, as described above.