Title :
Factoring solution sets of polynomial systems in parallel
Author :
Leykin, Anton ; Verschelde, Jan
Author_Institution :
Dept. of Math., Stat., & Comput. Sci., Univ. of Illinois at Chicago, IL, USA
Abstract :
The authors reported on a first parallel implementation of a recent algorithm to factor positive dimensional solution sets of polynomial systems. As the algorithm uses homotopy continuation, a good speedup of the path tracking jobs was observed. However, for solution sets of high degree, the overhead of managing different homotopies and large lists of solutions exposes the limits of the master/servant parallel programming paradigm for this type of problem. A probabilistic complexity study suggests modifications to the method, which will also improve the serial version of the original algorithm.
Keywords :
mathematics computing; parallel algorithms; parallel programming; polynomials; factoring solution sets; homotopy continuation; parallel polynomial systems; parallel programming; probabilistic complexity; Assembly systems; Computational geometry; Computer science; Concurrent computing; Equations; Mathematics; Parallel programming; Polynomials; Statistics; Uniform resource locators; 68W30; Linear trace; Primary 65H10; Secondary 14Q99; monodromy; numerical algebraic geometry; numerical homotopy algorithms; numerical irreducible decomposition; parallel computation; path following; polynomial systems;
Conference_Titel :
Parallel Processing, 2005. ICPP 2005 Workshops. International Conference Workshops on
Print_ISBN :
0-7695-2381-1
DOI :
10.1109/ICPPW.2005.31