Higher order differential analysis with vectorized derivatives (2011.01833v2)

Abstract: Higher order derivatives of functions are structured high dimensional objects which lend themselves to many alternative representations, with the most popular being multi-index, matrix and tensor representations. The choice between them depends on the desired analysis since each presents its own advantages and disadvantages. In this paper, we highlight a vectorized representation, in which higher order derivatives are expressed as vectors. This allows us to construct an elegant and rigorous algebra of vector-valued functions of vector variables, which would be unwieldy, if not impossible, to do so using the other representations. The fundamental results that we establish for this algebra of differentials are the identification theorems, with concise existence and uniqueness properties, between differentials and derivatives of an arbitrary order. From these fundamental identifications, we develop further analytic tools, including a Leibniz rule for the product of functions, and a chain rule (Fa`a di Bruno's formula) for the composition of functions. To complete our exposition, we illustrate how existing tools (such as Taylor's theorem) can be incorporated into and generalized within our framework of higher order differential calculus.