An adaptive algorithm for weighted approximation of singular functions over \(\mathbb{R}\) (Q2845599)

Given piecewise smooth univariate functions on the real line whose piecewise smoothness is of order \(r\), weighted approximations in the \(L^p\) context are studied where \(p\in[1,\infty]\). The points at which the approximand is evaluated are not fixed, but they are chosen adaptively depending on the function values of \(f\) at previous points. From this, one obtains a sequence of approximation operators with asymptotic accuracy \(n^{-r}\) where \(n\) is the index of the sequence of approximation operators and \(r\) is piecewise smoothness as mentioned above. Here, of course, there are conditions (necessary and sufficient) needed on the aforementioned weight functions, indeed their \(L^{1/\gamma}\) norm has to be finite where \(r=r+1/p\). This is necessary and -- under the condition that \(f\) is globally of smoothness \(r\) -- it is also sufficient.