Hardy-Littlewood-Paley inequality and recovery of derivatives from inaccurate data (Q656336): Difference between revisions

The authors study the Hardy-Littlewood-Paley inequality \[ \mid\mid x^{(k)}(.)\mid\mid_{L_{2}(\Re)} \leq \mid\mid x(.)\mid\mid^{1-k/n}_{L_{2}(\Re)} \mid\mid x^{(n)}(.)\mid\mid^{k/n}_{L_{2}(\Re)}, \] where \(x(.)\in {\mathcal W}^{n}_{2}(\Re),\) \(x^{(n)}(.)\in L_{2}(\Re)\), \(k\) and \(n\) are integers such that \(k<n\). They show that the sharpness of the above inequality means that for \(\delta_{1}>0\) and \( \delta_{2}>0\) the value of the problem \[ \mid\mid x^{(k)}(.)\mid\mid_{L_{2}(\Re)} \to \max\left\{\mid\mid x(.)\mid\mid_{L_{2}(\Re)}\leq \delta_ 1, \mid\mid x^{(n)}(.)\mid\mid_{L_{2}(\Re)}\leq \delta_ 2 \right\} \] is \( \delta^{1-k/n}_ 1\delta^{k/n}_ 2.\) The above problem is related to the optimal recovery of the \(k\)th derivative of the function \(x(.)\in {\mathcal W}^{n}_{2}(\Re)\) from inaccurate data on the function and its \(n\)th derivative. It is shown that the optimal recovery problem is equivalent to finding the error of optimal recovery \(E(D^{k}, {\mathcal W}^{n}_{2}(\Re), \delta_{1}, \delta_{2}).\) Specifically, it is proved that \[ E(D^{k}, {\mathcal W}^{n}_{2}(\Re), \delta_{1}, \delta_{2})= \delta^{1-k/n}_{1}\delta^{k/n}_{2}. \]