Content deleted Content added
m References |
Link suggestions feature: 3 links added. |
||
| (One intermediate revision by one other user not shown) | |||
Line 1:
'''Numerical certification''' is the process of verifying the correctness of a candidate solution to a [[system of equations]]. In (numerical) computational mathematics, such as [[numerical algebraic geometry]], candidate solutions are computed algorithmically, but there is the possibility that errors have corrupted the candidates. For instance, in addition to the inexactness of input data and candidate solutions, numerical errors or errors in the discretization of the problem may result in corrupted candidate solutions. The goal of numerical certification is to provide a certificate which proves which of these candidates are, indeed, approximate solutions.
Methods for certification can be divided into two flavors: ''a priori'' certification and ''a posteriori'' certification. ''A posteriori'' certification confirms the correctness of the final answers (regardless of how they are generated), while ''a priori'' certification confirms the correctness of each step of a specific computation. A typical example of ''a posteriori'' certification is [[Stephen Smale|Smale]]'s alpha theory, while a typical example of ''a priori'' certification is [[interval arithmetic]].
Line 48:
====Krawczyck method====
Let <math>Y</math> be any <math>n\times n</math> [[invertible matrix]] in <math>GL(n,\mathbb{R})</math>. Typically, one takes <math>Y</math> to be an approximation to <math>F'(y)^{-1}</math>. Then, define the function <math>G(x)=x-YF(x).</math> We observe that <math>x</math> is a fixed of <math>G</math> if and only if <math>x</math> is a root of <math>F</math>. Therefore the approach above can be used to identify roots of <math>F</math>. This approach is similar to a multivariate version of Newton's method, replacing the derivative with the fixed matrix <math>Y</math>.
We observe that if <math>J</math> is a compact and convex region and <math>y\in J</math>, then, for any <math>x\in J</math>, there exist <math>c_1,\dots,c_n\in J</math> such that
Line 56:
where the calculations are, once again, computed using interval arithmetic. Combining this with the formula for <math>G</math>, the result is the Krawczyck operator
:<math>K_{y,Y}(J)=y-YF(y)+(I-F'(J))(J-y),</math>
where <math>I</math> is the [[identity matrix]].
If <math>K_{y,Y}(J)\subset J</math>, then <math>G</math> has a fixed point in <math>J</math>, i.e., <math>F</math> has a root in <math>J</math>. On the other hand, if the maximum [[matrix norm]] using the [[Uniform norm|supremum norm for vectors]] of all matrices in <math>I-F'(J)</math> is less than <math>1</math>, then <math>G</math> is contractive within <math>J</math>, so <math>G</math> has a unique fixed point.
Line 81:
{{Main|Condition number}}
[[Numerical algebraic geometry]] solves polynomial systems using [[homotopy continuation]] and path tracking methods. By monitoring the [[condition number]] for a tracked homotopy at every step, and ensuring that no two solution paths ever intersect, one can compute a numerical certificate along with a solution. This scheme is called ''a priori'' path tracking.<ref>{{cite journal |last1=Beltran |first1=Carlos |last2=Leykin |first2=Anton |title=Certified numerical homotopy tracking |journal=Experimental Mathematics |date=2012 |volume=21 |issue=1 |pages=69–83}}</ref>
Non-certified numerical path tracking relies on heuristic methods for controlling time step size and precision.<ref>{{cite journal |last1=Bates |first1=Daniel |last2=Hauenstein |first2=Jonathan |last3=Sommese |first3=Andrew |last4=Wampler |first4=Charles |title=Stepsize control for path tracking |journal=Contemporary Mathematics |date=2009 |volume=496 |issue=21}}</ref> In contrast, ''a priori'' certified path tracking goes beyond heuristics to provide step size control that ''guarantees'' that for every step along the path, the current point is within the ___domain of [[quadratic convergence]] for the current path.
| |||