A unified representation for some interpolation formulas (Q2226664)

Both interpolation and quasi-interpolation in one or more dimension are useful, state-of-the-art approximation tools when continuous and smooth functions are approximated. There are a variety of subclasses, such as cardinal interpolation, meshless interpolation and quasi-interpolation etc. Also, approximants ([quasi-]interpolants) may be polynomials, splines (piecewise polynomials), kernel functions, radial basis functions; in other words there are many approximation spaces to start from. It is important to formulate (i.e., expand) interpolants in different ways for different applications, and the authors of this paper choose the Lagrange form to express the approximation. Unlike quasi-interpolation, interpolants may not always exist (or may not exist uniquely), depending mostly on the geometry of knots or interpolation points, but sometimes also on the space dimension. That is, the dimension of the underlying (mostly real) multidimensional space may be relevant, or the dimension of the approximation space from which the interpolants stem, or the knot placing in the case of piecewise polynomials (splines) or centres for radial basis functions. In the article, a special emphasis is placed on the mentioned existence and uniqueness, and a variety of classes of approximants and Lagrange functions is studied with examples.