Existence of continuous right inverses to linear mappings in finite-dimensional geometry (Q2663646)

If \(f: \mathbb{R}^n \to \mathbb{R}^m\) is a linear mapping, \(A \subseteq \mathbb{R}^n\) is a non-empty subset and \(B = f(A)\), the authors study the problem of the existence of a continuous right inverse of the mapping \(f |_A: A \to B\). Section 2 contains examples of \((f,A,B)\) for which the restriction \(f |_A : A \to B\) does not admit a continuous right inverse, with \(A\) compact and convex (in the third example in that section, which is a bit more complex than the first two examples, \(A\) has even smooth boundary). The authors then study sufficient conditions under which \(f |_A\) has a continuous right inverse. Without loss of generality, it is enough to consider the case where \(m = n-1\) and \(f\) is the projection mapping \[ (x_1, \ldots, x_n) \mapsto (x_1, \ldots, x_{n-1}), \] essentially because a linear projection from \(\mathbb{R}^n\) onto a subspace of codimension \(k\) can be obtained as the composition of \(k\) such projections onto a codimension \(1\) subspace at every step. Given a non-empty subset \(A \subseteq \mathbb{R}^m \times \mathbb{R}\), the authors define two functions \(u_A\) and \(w_A: \mathbb{R}^m \to [-\infty, \infty]\) by \begin{align*} u_A(x) &= \operatorname{inf}_{t \in \mathbb{R}} \{ t \in \mathbb{R}: (x,t) \in A \} \\ w_A(x) &= \operatorname{sup}_{t \in \mathbb{R}} \{ t \in \mathbb{R}: (x,t) \in A \} \end{align*} Note that \(u_A\) and \(w_A\) are not necessarily continuous. If one assumes that \(f^{-1}(x)\) is connected, for any \(x \in B\), then one can see that the problem of the existence of a continuous right inverse for \(f |_A\) is closely related to the problem of the existence of a continuous function \(w\) whose values are squeezed between \(u_A\) and \(w_A\). Sufficient conditions for the existence of a continuous right inverse are obtained in Sections 4, 5 and 6. For instance, Corollary 2 in Section 6 states the following. Corollary 2 Let \(A\) be a compact convex subset of \(\mathbb{R}^m \times \mathbb{R}\), let \(f\) be the restriction to \(A\) of the projection \(\mathbb{R}^m \times \mathbb{R}\) onto its first factor \(\mathbb{R}^m\), thus mapping \((x,t) \mapsto x\) and let \(B = f(A)\). Assume there exist two functions \(u, w: B \to [-\infty, \infty]\) such that \[ A = \{ (x,t) \in \mathbb{R}^m \times \mathbb{R}: u(x) \leq t \leq w(x) \}. \] Then \(f\) admits a continuous right inverse iff \(u^\sharp \leq w^\flat\) on \(\partial B\), where \(f^\sharp\) (\(f^\flat\)) denotes the upper (lower) regularization of a function \(f: B \to [-\infty, \infty]\). In Section 7, the authors consider the special case of polyhedra, and in Section 9, they formulate and prove a result which shows that projections which do not have a continuous right inverse are exceptional (rather than generic).