Divide and conquer roadmap for algebraic sets

From MaRDI portal
Publication:464736



Abstract: Let mathrmR be a real closed field, and mathrmDsubsetmathrmR an ordered domain. We describe an algorithm that given as input a polynomial PinmathrmD[X1,ldots,Xk], and a finite set, mathcalA=p1,ldots,pm, of points contained in V=mathrmZer(P,mathrmRk) described by real univariate representations, computes a roadmap of V containing mathcalA. The complexity of the algorithm, measured by the number of arithmetic operations in mathrmD is bounded by left(sumi=1mDiO(log2(k))+1ight)(klog(k)d)O(klog2(k)), where d=mathrmdeg(P), and Di is the degree of the real univariate representation describing the point pi. The best previous algorithm for this problem had complexity mathrmcard(mathcalA)O(1)dO(k3/2) due to Basu, Roy, Safey-El-Din, and Schost (2012), where it is assumed that the degrees of the polynomials appearing in the representations of the points in mathcalA are bounded by dO(k). As an application of our result we prove that for any real algebraic subset V of mathbbRk defined by a polynomial of degree d, any connected component C of V contained in the unit ball, and any two points of C, there exist a semi-algebraic path connecting them in C, of length at most (klog(k)d)O(klog(k)), consisting of at most (klog(k)d)O(klog(k)) curve segments of degrees bounded by (klog(k)d)O(klog(k)). While it was known previously, by a result of D'Acunto and Kurdyka, that there always exists a path of length (O(d))k−1 connecting two such points, there was no upper bound on the complexity of such a path.


In this paper, the authors deal with the following algorithmic problems: {\parindent=6mm \begin{itemize}\item[(1)] determining the number of semi-algebraically connected components of the set of zeroes of a multivariate polynomial in a real closed field; \item[(2)] decide whether two points belong to the same semi-algebraically connected component of this set; \item[(3)] if so, to compute a semi-algebraic path with image contained in this set. \end{itemize}} The main results of this paper are the following: Given and ordered subdomain \(D\) of a real closed field \(R\), there exists an algorithm that takes as input {\parindent=6mm \begin{itemize}\item[(1)] a polynomial \(P\in D[X_1,\dots, X_k]\), with \(\deg(P)\leq d\); \item[(2)] a finite set \({\mathcal A}=\{p_1,\dots, p_m\}\) contained in the ``real variety \(V\) defined by the zeroes of \(P\) in \(R^k\), such that for each \(i=1,\dots, m\), the degree of a ``real univariate representation (a specific algebraic way of representing these points with coefficients in \(D\)) of \(p_i\) is bounded by \(D_i\); \end{itemize}} and computes the following outputs: {\parindent=6mm \begin{itemize}\item[(1)] a ``roadmap (a semi-algebraic set of \(V\) of dimension at most one with some extra conditions) of \(V\) containing \({\mathcal A}\). \item[(2)] the number of semi-algebraically connected components of \(V\). \end{itemize}} The complexity of computing the first output is bounded by \[ \left(1+\sum_{i=1}^mD_i^{{\mathcal O}(\log^2k)}\right)\big(k^{\log k}d\big)^{{\mathcal O}(k\log^2k)}, \] while the size of the output is bounded by \(\big(\#{\mathcal A}+1\big)(k^{\log\,k}d)^{{\mathcal O}(k\,\log k)}\), and the degrees of the polynomials appearing in the descriptions of it are bounded by \(\big(\max_{1\leq i\leq m}D_i\big)^{{\mathcal O}(\log k)}\big(k^{\log k}d\big)^{{\mathcal O}(k\,\log k)}\). The complexity of computing the second output is bounded by \(\big(k^{\log k}d\big)^{{\mathcal O}(k\,\log^2 k)}\). If in addition we are given with the input two real univariate representations whose associated points are contained in \(V\) with degrees bounded by \(D_1\) and \(D_2\) respectively, there is an algorithm deciding whether the two points belong to the same semi-algebraically connected component of \(V\), and if so computes a description of a semi-algebraic path connecting them with image contained in \(V\). The complexity of this algorithm is bounded by \[ \left(D_1^{{\mathcal O}(\log^2k)}+D_2^{{\mathcal O}(\log^2k)}+1\right)\big(k^{\log k}d\big)^{{\mathcal O}(k \log^2k)}, \] and the size the the output as well as the degrees of the polynomials appearing in the descriptions of the curve segments and points are bounded by \(\max\{1,D_1,D_2\}^{{\mathcal O}(\log k)}\big(k^{\log k}d\big)^{{\mathcal O}(k \log k)}\). All these bounds above improve the previous known estimations, some of them gotten by the authors and some collaborators in [\textit{S. Basu} et al., Found. Comput. Math. 14, No. 6, 1117--1172 (2014; Zbl 1322.14090)]. In addition, as an application of the results above, the author prove the following result which was already addressed in [\textit{D. D'Acunto} and \textit{K. Kurdyka}, Bull. Lond. Math. Soc. 38, No. 6, 951--965 (2006; Zbl 1106.14046)] without upper bounds on complexity: if \(V\) is a real algebraic variety defined by a polynomial of degree at most \(D\), and \(C\) a connected component of \(V\) contained in the unit ball centered at the origin, any two points \(x, y\in C\) can be connected inside \(C\) by a semi-algebraic path of length at most \(\big(k^{\log k}d\big)^{{\mathcal O}(k \log k)}\), consisting of at most \(\big(k^{\log k}d\big)^{{\mathcal O}(k \log k)}\)curve segments of degrees bounded by \(\big(k^{\log k}d\big)^{{\mathcal O}(k \log k)}\).



Cites work


Cited in
(23)








This page was built for publication: Divide and conquer roadmap for algebraic sets

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q464736)