Constructive approach to the monotone rearrangement of functions (Q2121050)

A well-known exercise in analysis is to take a continuous function \(f:\Omega\to\mathbb{R}\) whose domain is a subset of finite measure of some \(\mathbb{R}^d\) and to look at the function \(f^\dag:(0,1)\to\mathbb{R}\) defined by \(f^\dag(y)=\inf\{u:\lambda[f\ge u]\ge y\cdot\lambda(\Omega)\}\), where \(\lambda\)~is Lebesgue measure and \([f\ge u]\) is short for \(\{x:f(x)\ge u\}\). Under suitable conditions one obtains that \(\int_0^1 f^\dag(y)\,\mathrm{d}y\) evaluates to \(\lambda(\Omega)^{-1}\cdot\int_\Omega f(x)\,\mathrm{d}\lambda\). The function \(f^\dag\) is the \textit{monotone rearrangement} of \(f\) and the authors prove that an oft-used way to approximate \(f^\dag\) is actually correct. The method is to enclose \(\Omega\) in a block, fill the block with a grid, sort the values of \(f\) on the grid in non-decreasing order and interpolate these values over a uniform partition in \((0,1)\) in a piecewise linear fashion. As the mesh of the grid goes to zero the sequence of interpolations will converge to \(f^\dag\). The paper relates the kind of convergence -- uniform, almost everywhere, etc. -- to regularity properties of \(f\) -- continuity, measurability, etc. -- and~\(\Omega\).