Modified iterated Tikhonov methods for solving systems of nonlinear ill-posed equations (Q534483): Difference between revisions

The article deals with the iterative Tikhonov--Kaczmarz algorithm (iTK-algorithm) for approximately solving an ill-posed system of nonlinear equations \[ F_i(x) = y_i, \qquad i = 0,1,\dots,N - 1,\tag{1} \] where \(F_i:\;D \subset X \to Y\) are operators between Hilbert spaces \(X\) and \(Y\). The iTK-algorithm is described by formulas \[ x_{k+1}^\delta = x_k^\delta - \alpha^{-1} F_{[k]}'(x_{k+1}^\delta)^*\left(F_{[k]}(x_{k+1}^\delta) - y_{[k]}^\delta\right)\tag{2} \] (\([k] := k \mod N\), \(x_0^\delta = x_0\)). It is assumed that the operators \(F_i\) are weakly sequentially continuous and Fréchet differentiable and \(\|F_j'(x)\| \leq M\) for \(x \in B_\rho(x_0)\) with \(M > 0\) and \(\rho > 0\); moreover, that the following local tangent cone condition \[ \left\|F_i(x) - F_i(\bar{x}) - F_i(\bar{x})(x - \bar{x})\right\|_Y \leq \eta\left\|F_i(x) - F_i(\bar{x})\right\|_Y, \quad x,\bar{x} \in B_\rho(x_0), \] holds for some \(\eta < 1\); and, at last, in \(B_{\rho/4}(x_0)\), there exists an exact solution \(x^*\) to the equation \(F(x) = y\), where \(y = (y_0,y_1,\ldots,y_{N-1})\) is the exact data in (1). It is proved that the algorithm (2) is a convergent regularization method. Further, the analogous result is proved for the iTK-method defined by \[ \begin{gathered} x_{k+1}^\delta = x_k^\delta - \alpha^{-1} \omega_k F_{[k]}'(x_{k+1}^\delta)^*\left(F_{[k]}(x_{k+1}^\delta) - y_{[k]}^\delta|\right),\\ \omega_k := \begin{cases} 1 & \|F_{[k]}(x_{k+1}^\delta) - y_{[k]}^\delta\| \geq \tau\delta_{[k]}, \\ 0 & \text{otherwise}. \end{cases} \end{gathered} \] At the end of the article some discussion about possible applications related to parameter identification in elliptic partial differential equations is presented.