Implicit representations of high-codimension varieties (Q2010333)

Geometric objects can be represented for CAGD and other applications in more than one way. A common and useful representation is a parametric one, where the parameter values determine the location in space of the corresponding piece of the object. Another useful representation is a so-called \textit{implicit} representation which is usually presented as a set of multivariate polynomial generators, implicitly describing the object by the values of the coordinates that make these generators equal to zero. Deciding whether a given point is \textit{inside} or \textit{outside} an object can be easier with an implicit representation, for instance. Taking a parametric representation of an object and producing an implicit one is called \textit{implicitization}. This paper presents a new randomized algorithm, claimed to be robust and efficient, for implicitizing space curves and varieties of codimension higher than one. The paper uses the theory of Chow forms to obtain the equations of conical hypersurfaces intersecting at precisely the desired geometric object. The paper gives a readable introduction to Chow forms (with references) in their Section 2. The paper gets quite deep very quickly; the authors are expert. The paper also uses \textit{implicitization matrices}, and provides algorithmic descriptions of its main results. The major bottleneck, they claim, is the computation of multivariate resultants. Undoubtedly this is true. This is even more true because of a recent result (not cited in the paper) which says that multivariate resultants are inherently ill-conditioned: see [\textit{V. Noferini} and \textit{A. Townsend}, SIAM J. Numer. Anal. 54, No. 2, 719--743 (2016; Zbl 1382.65144)]. Nonetheless, in exact arithmetic and with exactly-known data, the main algorithm of the paper being reviewed seems like it will be useful.