Relationships of properties of piecewise affine maps over ordered fields (Q919066)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Relationships of properties of piecewise affine maps over ordered fields |
scientific article |
Statements
Relationships of properties of piecewise affine maps over ordered fields (English)
0 references
1990
0 references
Let G be an ordered field, and let \(G^{m\times n}\) denote matrices with elements in G. In the G-vector space \(G^ n=G^{n\times 1}\) the norm \(\| \|_{\infty}\) defined by the absolute value in G induces a topology on \(G^ n\). A set \(\sigma \subseteq G^ n\) is called a cell if it can be represented as \(\{x\in G^ n:\) Ax\(\leq a\}\) for some \(A\in G^{p\times n}\) and \(a\in G^ n\). Then (p,A,a) is refered as a representation of \(\sigma\). A subset \(\tau\) of a cell \(\sigma\) is called a face of \(\sigma\), if \(\tau =\emptyset\), or \(\tau =\sigma\), or \(\tau =\{x\in \sigma:\) \(A_{\alpha}x\leq a_{\alpha}\}\) for some subsequence \(\alpha\) of 1,2,...,n, where \(A_{\alpha}\) and \(a_{\alpha}\) are corresponding submatrices. A finite collection \({\mathcal S}\) of cells in \(R^ n\) is said to be a subdivision of \(G^ n\) if (a) \({\mathcal S}\) covers \(G^ n\), (b) each face of each cell in \({\mathcal S}\) is also in \({\mathcal S}\), (c) each cell in \({\mathcal S}\) is a face of some cell in \({\mathcal S}^ n\), the set of all cells in \({\mathcal S}\) of dimension n, and (d) the intersection of any two cells of \({\mathcal S}\) is a face, perhaps empty, of both cells. If \((p_{\sigma},A_{\sigma},a_{\sigma})\) is a representation of \(\sigma\), then \(((p_{\sigma},A_{\sigma},a_{\sigma}):\sigma\in {\mathcal S})\) is refered as a representation of \({\mathcal S}.\) A function f: \(G^ n\to G^ k\) is defined to be piecewise affine with respect to \({\mathcal S}\), or \({\mathcal S}-PA\), if f is affine on each cell \(\sigma\) in \({\mathcal S}\), i.e. if \(f(x)=B_{\sigma}x+b_{\sigma}\) for all \(x\in \sigma\). In this case, \(((B_{\sigma},b_{\sigma}):\sigma\in {\mathcal S}^ n)\) is refered as \({\mathcal S}\)-representation of f. A function f: \(G^ n\to G^ k\) is said to be PA map if f is \({\mathcal S}-PA\) for some subdivision \({\mathcal S}\) of \(G^ n.\) In this paper, the authors study the relationships between some properties of PA maps as local and global variation of univalence, openness, onto condition, nonexpansiveness, and orientation and eigenvalue conditions on derivatives. Using two general transfer principles, they extend a subdivision \({\mathcal S}\) of \(G^ n\) to a subdivision \({\mathcal S}_ H\) of \(H^ n_ n\) with the same representation, and then define PA extension \(f_ H:\) \(H^ n\to H^ k\) of a PA map f: \(G^ n_ n\to G^ k\) by an \(S_ H\)-representation corresponding to \({\mathcal S}\)-representation of f. Using these principles, the authors show that most of the mentioned properties of f are preserved at the maps \(f_ H\) are extended to larger ordered fields H.
0 references
piecewise affine maps
0 references
ordered field
0 references
cell
0 references
face
0 references
subdivision
0 references
representation
0 references
univalence
0 references
openness
0 references
onto condition
0 references
nonexpansiveness
0 references
orientation
0 references
eigenvalue conditions
0 references
0 references