On the characterization and computation of best monotone approximation in \(L_ p[0,1]\) for \(1\leq p<\infty\) (Q908455)

Let \(M_ p\subset L_ p[0,1]\) denote the closed convex lattice of nondecreasing functions in \(L_ p\). Then \(g^*\in M_ p\) is called a best nondecreasing (nd) \(L_ p\) approximation to \(f\in L_ p\) if and only if \(\| f-g^*\|_ p\leq \| f-g\|_ p\) for all \(g\in M_ p\). For \(1\leq p<\infty\) each \(f\in L_ p\) has a best nd \(L_ p\) approximation which is unique for \(1<p<\infty\) (up to sets of Lebesgue measure zero). Constructive solutions to this approximation problem are known. In the present paper the authors give an alternative characterization of such best approximations. They use it to develop algorithms for the computation of best nd \(L_ p\) approximation.