A note on a theorem of Prolla (Q1194641)

Let \(C\) be a nonempty, closed, convex subset of a normed space \(X\). A map \(g: C\to X\) is called almost affine if \(\| gz-y\|\leq\lambda\| gz_ 1-y\|_ (1-\lambda)\| gx_ 2-y\|\), for \(x_ 1,x_ 2\in C\), \(0<\lambda<1\), \(z=\lambda x_ 1+(1-\lambda)x_ 2\) and \(y\in X\). If \(\| gz-y\|=\max(\| gx_ 1-y\|,\| gx_ 2-y\|)\), then \(g\) is called almost quasi-convex. The main result of this paper asserts that if \(C\) is a nonempty, compact, convex subset of a normed space \(X\) and \(g: C\to C\) is a continuous, almost quasi-convex, surjective mapping, then for every continuous function \(f: C\to X\) there exists \(y\in C\) such that \(\| gy-fy\|=d(fy,C)\). If \(fy\in C\) one obtains a coincidence result. For \(fy\in C\) and \(g\) the identity map one obtains a fixed point theorem and for \(g\) the identity map and \(fy=x\), for all \(y\in C\), one obtains a best approximation result. In the case of an almost quasi- affine function the theorem was proved by \textit{J. B. Prolla} [Numer. Funct. Anal. Optimization 9, 619-628 (1978)]. In Theorem 2, a similar result is obtained with some relaxation on the compactness of \(C\).