A right-preconditioning process for the formal-algebraic approach to inner and outer estimation of AE-solution sets. (Q2568634)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A right-preconditioning process for the formal-algebraic approach to inner and outer estimation of AE-solution sets. |
scientific article |
Statements
A right-preconditioning process for the formal-algebraic approach to inner and outer estimation of AE-solution sets. (English)
0 references
18 October 2005
0 references
Let \({\mathbf A}= ({\mathbf a}_{ij})\) be an \(n\times n\) interval matrix and \({\mathbf b}= ({\mathbf b}_k)\) be an interval vector with \(n\) cmponents. Assume that with each interval \({\mathbf a}_{ij}\), \({\mathbf b}_k\) there is associated a quantifier \(\alpha_{ij}\), \(\beta_k\in \{\forall,\exists\}\). Denote by \(\{\gamma_k\mid k=1,\dots, p\}\), \(\{\overline\gamma_k\mid k=1,\dots, q\}\) and \(\{\delta_k\mid k= 1,\dots, s\}\), \(\{\overline\delta_k\mid k= 1,\dots, t\}\), respectively, the set of indices \((i, j)\) belonging to \(\alpha_{ij}= \forall\), \(\alpha_{ij}=\exists\), and set of indices \(k\) belonging to \(\beta_k= \forall\), \(\beta+k= \exists\), respectively. Then the AE-solution set \(\Xi_{\alpha\beta}({\mathbf A},{\mathbf b})\) is defined by \[ \begin{multlined} \Xi_{\alpha\beta}({\mathbf A},{\mathbf b})= \{x\in\mathbb{R}^n\mid (\forall a_{\gamma_1}\in{\mathbf a}_{\gamma_1})\cdots(\forall a_{\gamma_p}\in{\mathbf a}_{\gamma_p})(\forall b_{\delta_1}\in{\mathbf b}_{\delta_1})\cdots(\forall b_{\delta_s}\in{\mathbf b}_{\delta_s})\\ (\exists a_{\overline\gamma_1}\in{\mathbf a}_{\overline\gamma_1})\cdots(\exists a_{\overline\gamma_q}\in {\mathbf a}_{\overline\gamma_q})(\exists b_{\overline\delta_1}\cdots (\exists b_{\overline\delta_t}\in{\mathbf b}_{\overline\delta_t})(Ax= b\}.\end{multlined} \] With \({\mathbf A}\), \({\mathbf b}\), \((\alpha_{ij})\), \((\beta_k)\) there are uniquely associated some characteristic interval matrix \(\widetilde{\mathbf A}\) and some characteristic interval vector \(\widetilde{\mathbf b}\) with entries which may be improper intervals, i.e., intervals \([\underline c, \overline c]\) with \(\underline c\geq \overline c\). Using an appropriate extended interval arithmetic auxiliary interval equations are derived for \(\widetilde{\mathbf A}\), \(\widetilde{\mathbf b}\) whose algebraic solutions \({\mathbf x}= [\underline x, \overline x]\) are inner or outer estimates of \(\Xi_{\alpha\beta}({\mathbf A}, {\mathbf b})\) if \(\underline x\leq\overline x\). Replacing \(\widetilde{\mathbf A}\) by the matric product \(\widetilde{\mathbf A}C\) (= right-preconditioning) with some invertible real \(n\times n\) matrix \(C\) and solving the systems for the corresponding characteristic quantities yields inner and outer enclosures of \(\Xi_{\alpha\beta}({\mathbf A},{\mathbf b})\) in the form of parallelepipeds, provided that appropriate additional conditions hold. For \(C\) being the midpoint inverse of \({\mathbf A}\) results on the quality of these enclosures are given. In some cases they turn out to be superior to those being discussed in the literature up to now. Numerical experiments conclude the paper.
0 references
AE-solution sets
0 references
right-preconditioning
0 references
inner enclosure
0 references
outer enclosure
0 references
interval linear system
0 references
generalized intervals
0 references
improper interval
0 references
single step method
0 references
numerical experiments
0 references
0 references