A comparison of error estimates at a point and on a set when solving ill-posed problems (Q1654225)

The following problem is considered. It is supposed that for a function \(f_0 \in L_2(-\infty,\infty)\) the function \(u_0 =Tf_0 \in Z\) is known, where \(Tf(\xi) = t(\xi)f(\xi)\) and \(f(\xi) \in \{f(\xi) \: : \: f(\xi) \in L_2(-\infty,\infty) , \,t(\xi)f(\xi) \in L_2(-\infty,\infty)\}\). The function \(t(\xi)\) is an even function from the space \(C(-\infty,\infty)\) with \(t(\xi) \rightarrow \infty\) when \(\xi \rightarrow \infty\), \(t(0) > 0\), and \(t(\xi)\) is strictly increasing on \([0,\infty)\). \(Z\) denotes the set of Fourier transform extensions of a class of piecewise smooth functions. Instead of the exact (unknown) function \(f_0\), a function \(f_\delta \in L_2(-\infty,\infty)\) with \(\| f_\delta - f_0\| \leq \delta\) is given. An approximate solution \(u_\delta \in L_2(-\infty,\infty)\) and estimates of \(\| u_\delta - u_0 \|\) are to be determined. The nonlinear projective regularization method (see, e.g., [\textit{V. P. Tanana}, Sib. Zh. Ind. Mat. 5, No. 4, 150--163 (2002; Zbl 1053.65043)]) is used for solving this problem. A relation between the error at the point \(u_0 \in Z\) and the error on the correctness class is proved. Finally, the presented approach is demonstrated by a Cauchy inverse problem for the heat transfer equation.