Approximate varieties, approximate ideals and dimension reduction (Q2458336): Difference between revisions

The author discusses a method to compute real algebraic varieties of comparatively low dimension and low degree which contain given points in a numerical, inexact setting. It relies on the notions of \(p\)-approximate ideal \(I_{p,\varepsilon}(\Xi)\) and approximate H-basis associated with a finite set \(\Xi\subset\mathbb R^d\). The former is defined by \(f\in I_{p,\varepsilon}(\Xi) \Leftrightarrow \| f(\Xi)\| _{\ell^p(\Xi)}\leq \varepsilon\| f\| \), where \(\| f\| ^2\) equals the squared sum of coefficients of \(f\), and \(p\in [1,\infty]\). The latter means an H-basis contained in \(I_{p,\varepsilon}(\Xi)\). The main result of the paper is a recursive procedure for the computation of an approximate H-basis, which involves a QR decomposition of Vandermonde matrices. Further, a rather pessimistic a priori error bound is given.