## Geometric Completion of Differential Systems using Numeric-Symbolic Continuation (2002)

Venue: | SIGSAM Bulletin |

Citations: | 10 - 8 self |

### BibTeX

@ARTICLE{Reid02geometriccompletion,

author = {Greg Reid and Chris Smith},

title = {Geometric Completion of Differential Systems using Numeric-Symbolic Continuation},

journal = {SIGSAM Bulletin},

year = {2002},

volume = {36},

pages = {1--17}

}

### OpenURL

### Abstract

Symbolic algorithms using a finite number of exact differentiations and eliminations are able to reduce over and under-determined systems of polynomially nonlinear differential equations to involutive form. The output involutive form enables the identification of consistent initial values, and eases the application of exact or numerical integration methods. Motivated to avoid expression swell of pure symbolic approaches and with the desire to handle systems with approximate coefficients, we propose the use of homotopy continuation methods to perform the differentialelimination process on such non-square systems. Examples such as the classic index 3 Pendulum illustrate the new procedure. Our approach uses slicing by random linear subspaces to intersect its jet components in finitely many points. Generation of enough generic points enables irreducible jet components of the differential system to be interpolated. 1

### Citations

268 |
Computer methods for ordinary differential equations and differential-algebraic equations
- Ascher, Petzold
- 1998
(Show Context)
Citation Context ...uare) systems of ODE arise in applications such as constrained multibody mechanics and control systems. For example, differential-algebraic equations (DAE) arise from constrained Lagrangian mechanics =-=[2, 28, 30]-=-. In this article we consider systems of constrained ODE which are polynomial functions of their dependent and independent variables. For example consider the class of DAE of the form ut ¡ f ¢ u£ , φ ... |

152 |
Algorithm 795: PHCpack: A general-purpose solver for polynomial systems by homotopy continuation
- Verschelde
- 1999
(Show Context)
Citation Context ...special class of projection operators. More efficient decomposition methods use monodromy [35] and interpolate on a structured grid of samples [36]. The implementation [37] as extra module to PHCpack =-=[43]-=- is capable of handling large systems such as the cyclic 8 and 9-roots problem, and several applications from mechanical engineering [39]. 4.1 A Visualizable Example To illustrate our numerical proced... |

113 | Computing representations for radicals of a finitely generated differential ideals. Available at http://www.lifl.fr/boulier/publications.html
- Boulier, Lazard, et al.
(Show Context)
Citation Context ...5739 and Grant No. 0134611. 1Or more explicitly d dt φ ¦ j ∑k ∂φ j ∂uk ∂uk ¦ ∂t 0. 1sGeometric Completion of Differential Systems Formally reviewed communication Recently there has been much progress =-=[4, 17, 19, 20, 22, 25, 27, 41, 44]-=- in the theoretical development and implementation of symbolic differential-elimination algorithms to locate all such constraints. The elimination phase can be complicated because the equations are ge... |

105 | Applied Analysis - Lanczos - 1957 |

75 | Continuation and path following
- Allgower, Georg
- 1993
(Show Context)
Citation Context ...¡ 0 and lead to the solutions of p ¢ x£ ¡ 0 as t approaches 1. When q has a structure similar to that of p (see [16] for precise choices of q), we are guaranteed that numerical path following methods =-=[1]-=- will lead to approximations to all isolated solutions of p ¢ x£ ¡ 0. The systems of polynomial equations involved in our approach are usually not square and generally have positive dimensional compon... |

68 |
Systems of Partial Differential Equations and Lie Pseudogroups
- Pommaret
- 1978
(Show Context)
Citation Context ...thods cannot handle approximate input. An underlying principle of our approach, is our strong emphasis on geometry. In particular we emphasize jet space geometry, the geometry of differential systems =-=[18, 22, 27, 42]-=-. In comparison to the symbolic differentiationelimination approaches, where symbolic or algebraic manipulations of the equations figure heavily, our approach focuses on the solutions of the system re... |

65 | Numerical decomposition of the solution sets of polynomial systems into irreducible components
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ... given systems in square polynomial systems, by the inclusion of extra (slack) variables and random linear equations if necessary. The key idea behind these methods of Sommese, Verschelde and Wampler =-=[33, 34, 35, 36, 37]-=- is to reduce to the zero dimensional case (where there are only finitely many solutions), by slicing the solutions with a random linear space of the appropriate dimension. Enough points are located t... |

60 |
Numerical solution of multivariate polynomial systems by homotopy continuation methods
- Li
- 1997
(Show Context)
Citation Context ...m 0 to 1. The paths x ¢ t£ start at t ¡ 0 at known solutions of the start system q ¢ x£ ¡ 0 and lead to the solutions of p ¢ x£ ¡ 0 as t approaches 1. When q has a structure similar to that of p (see =-=[16]-=- for precise choices of q), we are guaranteed that numerical path following methods [1] will lead to approximations to all isolated solutions of p ¢ x£ ¡ 0. The systems of polynomial equations involve... |

57 | Numerical homotopies to compute generic points on positive dimensional algebraic sets
- Sommese, Verschelde
(Show Context)
Citation Context ...omponent, and k is the largest dimension of any component. A general method to determine the top dimension of the solution set of a polynomial system is to start the sequence of homotopies defined in =-=[32]-=- at dimension ¥ n 1 (where n is the dimension of the ambient space) and to remove linear slices until generic points are found. For the specific systems arising in the treatment of DAE, more efficient... |

52 | Complexity of Bézout’s theorem V: Polynomial time
- Shub, Smale
- 1994
(Show Context)
Citation Context ...ies for improving the efficiency and underlying complexity of our method. Newton’s method has quadratic order of convergence, once it is in the basin of attraction of a root. Moreover, Shub and Smale =-=[31]-=- demonstrated that the problem of whether a generic system has a root, can be answered in polynomial time (under certain circumstances, which include the absence of multiplicities). For the homotopy m... |

45 | Reduction of systems of nonlinear partial dierential equations to simpli involutive forms
- Reid, Wittkopf, et al.
- 1996
(Show Context)
Citation Context ...e components have multiplicity greater than 1 is that the symbolic derivative of an equation, may not be equivalent to a geometric derivative of an equation. This phenomenon is discussed in detail in =-=[21, 22]-=-. Simply speaking, for the symbolic algorithm to be a faithful representation of the underlying geometric algorithm, the ideals generated by the intermediate systems should be radical. This is a gener... |

44 |
Differential Gröbner bases
- Mansfield
- 1991
(Show Context)
Citation Context ...5739 and Grant No. 0134611. 1Or more explicitly d dt φ ¦ j ∑k ∂φ j ∂uk ∂uk ¦ ∂t 0. 1sGeometric Completion of Differential Systems Formally reviewed communication Recently there has been much progress =-=[4, 17, 19, 20, 22, 25, 27, 41, 44]-=- in the theoretical development and implementation of symbolic differential-elimination algorithms to locate all such constraints. The elimination phase can be complicated because the equations are ge... |

43 | Symmetric functions applied to decomposing solution sets of polynomial systems
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ... given systems in square polynomial systems, by the inclusion of extra (slack) variables and random linear equations if necessary. The key idea behind these methods of Sommese, Verschelde and Wampler =-=[33, 34, 35, 36, 37]-=- is to reduce to the zero dimensional case (where there are only finitely many solutions), by slicing the solutions with a random linear space of the appropriate dimension. Enough points are located t... |

41 | Using monodromy to decompose solution sets of polynomial systems into irreducible components. In Application of Algebraic Geometry to Coding Theory, Physics and Computation, edited by
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ... given systems in square polynomial systems, by the inclusion of extra (slack) variables and random linear equations if necessary. The key idea behind these methods of Sommese, Verschelde and Wampler =-=[33, 34, 35, 36, 37]-=- is to reduce to the zero dimensional case (where there are only finitely many solutions), by slicing the solutions with a random linear space of the appropriate dimension. Enough points are located t... |

39 |
The index of general nonlinear DAEs
- Campbell, Gear
- 1995
(Show Context)
Citation Context ...g on the component. There has been considerable discussion about different definitions of indices, some non-geometric, some geometric and some involving perturbations of the solutions of such systems =-=[5, 10, 15, 29]-=-. The geometric definition of the index we use is bounded by the number m of dependent variables [22, 41]. It is important that geometric differentiations are used in the completion procedure. Symboli... |

31 |
A geometric treatment of implicit differential-algebraic equations
- Rabier, Rheinboldt
- 1994
(Show Context)
Citation Context ...on polynomials. 10s6.1 Indices G. Reid, C. Smith & J. Verschelde The differential index we use is geometric [22], and and is closely related to that introduced by Reich [24] and Rabier and Rheinboldt =-=[23]-=-. For systems with several components, it can take different values, depending on the component. There has been considerable discussion about different definitions of indices, some non-geometric, some... |

31 |
Analysis and Application of the Formal Theory of Partial Differential Equations
- Seiler
- 1994
(Show Context)
Citation Context ...5739 and Grant No. 0134611. 1Or more explicitly d dt φ ¦ j ∑k ∂φ j ∂uk ∂uk ¦ ∂t 0. 1sGeometric Completion of Differential Systems Formally reviewed communication Recently there has been much progress =-=[4, 17, 19, 20, 22, 25, 27, 41, 44]-=- in the theoretical development and implementation of symbolic differential-elimination algorithms to locate all such constraints. The elimination phase can be complicated because the equations are ge... |

28 |
Deformation techniques for efficient polynomial equation solving
- Heintz, Krick, et al.
(Show Context)
Citation Context ...symbolically represent a determinant as a sum of monomials versus more cheaply evaluating it numerically at a point. This line of research has been followed by Giusti, Heintz [9], their collaborators =-=[11]-=- and recently Lecerf [14]. They have obtained a probabilistic algorithm (the geometric resolution) which if the input system is successfully encoded as straight line programs, can solve systems with c... |

24 | Numerical Irreducible Decomposition using Projections from Points on the Components. Preprint 2000. Received 13 November, 2000 A. Calabri, Dipartimento di Matematica, UniversitaÁ di Roma ``Tor Vergata'', via della Ricerca Scienti®ca, 00133 Roma E-mail: ca
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ...reducible decomposition [37]. PHCpack contains symbolic differentiation methods for polynomials for the differentiations needed in this paper. PHCpack uses polynomials interpolating through projected =-=[34]-=- generic points to represent irreducible solution components [33]. The computation of these interpolation polynomials is one of the most expensive parts of the method, and can be the least well condit... |

23 | Numerical irreducible decomposition using PHCpack - Sommese, Verschelde, et al. - 2003 |

21 | Differential elimination-completion algorithms for
- Reid, Lin, et al.
(Show Context)
Citation Context ...systems being the easiest to solve numerically. This index was originally introduced by Gear and its definition was refined and made coordinate independent in [23, 24]. For algebraic developments see =-=[22, 41]-=-. This article was formally reviewed following the procedures described in this Bulletin, 32(2), issue 124, 1998, pp. 5-6. ¥ † Support from NSERC is gratefully acknowledged. ‡ This material is based u... |

19 | Factorization free decomposition algorithms in differential algebra
- Hubert
(Show Context)
Citation Context ...c algorithms to produce primary decompositions are notoriously complicated and difficult to implement (implementations do not exist in Mathematica and Maple). For partial differential systems, Hubert =-=[12]-=- has created factorization-free decomposition algorithms and Arponen has designed algorithms for analogues of primary decomposition for ordinary differential systems [3]. Any differential system becom... |

18 | Kronecker’s smart, little black boxes
- Giusti, Heinz
- 2001
(Show Context)
Citation Context ...sical illustration is to symbolically represent a determinant as a sum of monomials versus more cheaply evaluating it numerically at a point. This line of research has been followed by Giusti, Heintz =-=[9]-=-, their collaborators [11] and recently Lecerf [14]. They have obtained a probabilistic algorithm (the geometric resolution) which if the input system is successfully encoded as straight line programs... |

18 | Advances in polynomial continuation for solving problems in kinematics
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ...mples [36]. The implementation [37] as extra module to PHCpack [43] is capable of handling large systems such as the cyclic 8 and 9-roots problem, and several applications from mechanical engineering =-=[39]-=-. 4.1 A Visualizable Example To illustrate our numerical procedure we consider the non-physical example xt ¤ y 2 ¥ 4 ¡ 0sxt ¤ 2x 2 ¤ 3y 2 ¥ 6 ¡ 0 ¢ (11) 0. Square the System. The dimension of the cons... |

18 |
Numerical algebraic geometry, in The Mathematics of Numerical Analysis
- Sommese, Wampler
- 1996
(Show Context)
Citation Context ...are and generally have positive dimensional components (submanifolds) of solutions. Our treatment of such systems, using homotopy methods, is made possible by recent theoretical progress initiated by =-=[40]-=-, in the developing area of Numerical Algebraic Geometry. The new techniques rely on embedding the given systems in square polynomial systems, by the inclusion of extra (slack) variables and random li... |

17 | A probabilistic algorithm to test local algebraic observability in polynomial time
- Sedoglavic
(Show Context)
Citation Context ...on complicated examples. This explosion reflects the underlying non-polynomial complexity, an exception is the probabilistic method for determining algebraic observability in polynomial time given in =-=[26]-=-. Moreover, symbolic methods cannot handle approximate input. An underlying principle of our approach, is our strong emphasis on geometry. In particular we emphasize jet space geometry, the geometry o... |

17 | A method for tracking singular paths with application to the numerical irreducible decomposition
- Sommese, Verschelde, et al.
(Show Context)
Citation Context ...gh degree components, to use extra samples for better conditioning of the interpolants. Finally we mention that the difficulties of higher multiplicity components are not discussed in this paper, see =-=[38]-=- for recent progress. 6 Indices, complexity and systems of PDE The differential index gives a measure of complexity due to differentiation in the geometric completion of systems of DAE. We discuss the... |

16 | Involution and constrained dynamics I: The Dirac approach
- Seiler, Tucker
- 1995
(Show Context)
Citation Context ...uare) systems of ODE arise in applications such as constrained multibody mechanics and control systems. For example, differential-algebraic equations (DAE) arise from constrained Lagrangian mechanics =-=[2, 28, 30]-=-. In this article we consider systems of constrained ODE which are polynomial functions of their dependent and independent variables. For example consider the class of DAE of the form ut ¡ f ¢ u£ , φ ... |

15 |
A geometricnumeric algorithm for factoring multivariate polynomials
- Corless, Galligo, et al.
- 2002
(Show Context)
Citation Context ...nd equations. 5sGeometric Completion of Differential Systems Formally reviewed communication Decompose(F): This algorithm uses incremental interpolation [33], or monodromy as is proposed in [35] (see =-=[6]-=- for the application of the special case of factoring multivariate polynomials with approximate coefficients) to return ¢sk£ G where G are the generic points classified by component, and k is the larg... |

15 |
Une alternative aux méthodes de réécriture pour la résolution des systèmes algébriques. PhD thesis, École polytechnique
- Lecerf
- 2001
(Show Context)
Citation Context ...determinant as a sum of monomials versus more cheaply evaluating it numerically at a point. This line of research has been followed by Giusti, Heintz [9], their collaborators [11] and recently Lecerf =-=[14]-=-. They have obtained a probabilistic algorithm (the geometric resolution) which if the input system is successfully encoded as straight line programs, can solve systems with complexity that is a polyn... |

14 | Existence and uniqueness theorems for formal power series solutions of analytic dierential systems
- Rust, Reid, et al.
- 1999
(Show Context)
Citation Context |

13 | Fast Differential Elimination in C: The CDiffElim Environment
- Wittkopf, Reid
- 2001
(Show Context)
Citation Context ...e termination criteria. The usual termination tests which are based on the highest derivatives of a system do not always apply (see [45, Section 6, Example 4] for a case where they fail). As shown in =-=[45]-=- a projected version of those tests should be used to see when a projected system becomes involutive. Checking dimensions, and existence of components in these projected systems, can be done without i... |

11 | On the numerical solution of involutive ordinary differential systems: Higher order methods 2, Research Report A409 - Tuomela, Arponen - 1999 |

8 | Involution and constrained dynamics II: The Faddeev-Jackiw approach - Seiler - 1995 |

8 |
Numerical Solution of Differential Algebraic Equations, Global Error Estimation and Symbolic Index Reduction. Ph.D. Thesis. Laboratoire de Modélisation et Calcul
- Visconti
- 1999
(Show Context)
Citation Context |

7 |
Some remarks on solvability and various indices for implicit differential equations
- Vey
- 1998
(Show Context)
Citation Context ...g on the component. There has been considerable discussion about different definitions of indices, some non-geometric, some geometric and some involving perturbations of the solutions of such systems =-=[5, 10, 15, 29]-=-. The geometric definition of the index we use is bounded by the number m of dependent variables [22, 41]. It is important that geometric differentiations are used in the completion procedure. Symboli... |

6 | Indices and Solvability of General Systems of Differential equations
- Seiler
- 1999
(Show Context)
Citation Context ...g on the component. There has been considerable discussion about different definitions of indices, some non-geometric, some geometric and some involving perturbations of the solutions of such systems =-=[5, 10, 15, 29]-=-. The geometric definition of the index we use is bounded by the number m of dependent variables [22, 41]. It is important that geometric differentiations are used in the completion procedure. Symboli... |

4 |
Beitrag zur Theorie der Algebrodifferentialgleichungen
- Reich
- 1990
(Show Context)
Citation Context ...ng such systems, with low (0 or 1) index systems being the easiest to solve numerically. This index was originally introduced by Gear and its definition was refined and made coordinate independent in =-=[23, 24]-=-. For algebraic developments see [22, 41]. This article was formally reviewed following the procedures described in this Bulletin, 32(2), issue 124, 1998, pp. 5-6. ¥ † Support from NSERC is gratefully... |

4 |
Symbolic computation of the index of quasilinear differential-algebraic equations
- Thomas
- 1996
(Show Context)
Citation Context ...systems being the easiest to solve numerically. This index was originally introduced by Gear and its definition was refined and made coordinate independent in [23, 24]. For algebraic developments see =-=[22, 41]-=-. This article was formally reviewed following the procedures described in this Bulletin, 32(2), issue 124, 1998, pp. 5-6. ¥ † Support from NSERC is gratefully acknowledged. ‡ This material is based u... |

3 |
Perturbation versus differentiation indices
- Hausdorf, Seiler
- 2001
(Show Context)
Citation Context |

2 |
Naumann (eds.) Automatic Differentiation 2000: From Simulation to Optimization
- Corliss, Faure, et al.
- 2001
(Show Context)
Citation Context ...cal decomposition method can take as input polynomial systems encoded as straight line programs. Moreover Newton methods, at the core of this method, can use methods such as automatic differentiation =-=[7]-=- to differentiate the programs, and implement the homotopy methods. Junk points can be rejected by a homotopy (and thus can again use a straight line encoding). These observations, form the basis of o... |

1 |
The complete form of a differential algebraic equation
- Arponen
(Show Context)
Citation Context ...les in the system. Here J0 ¡sm and J1 ¡s2m , suppressing the t dependence. It should be noted that higher order DAE with explicit time dependence can be dealt with directly in the Jet space formalism =-=[3, 27, 42]-=- without recourse to the above transformations. This has the advantage of representing the same geometric object with fewer variables. For example, the pendulum example, can be executed more efficient... |

1 |
On initial value problems for ordinary DAEs
- Pritchard, Sit
(Show Context)
Citation Context |

1 | An algorithm for the reduction of linear DAE
- Quéré, Villard
- 1995
(Show Context)
Citation Context |