Exact support recovery for linear inverse problems with sparsity constraints (Q356671)

This paper concerns the linear inverse problems \(Ku=g^{\varepsilon},\) solution \(u \in \ell^2,\) and bounded injective operator \(K: \ell^2 \rightarrow \mathcal{H}\) between the sequence space \(\ell^2\) and a separable Hilbert space \(\mathcal{H},\) where the observation \(g^{\varepsilon}\) is noisy, i.e., \(g^{\varepsilon} = g+\mu \in \mathcal{H},\) \(\|g-g^{\varepsilon}\|=\|\mu\|\leq \varepsilon\). The operator equation \(Ku=g\) is assumed to has a finitely supported solution \(u^*\). Conditions of exact recovery of the support of \(u^*\) are formulated and valid both for \(\ell^1\)-Tikhonov regularization and for the recently proposed sparsity enforcing orthogonal matching pursuit method (cf.\ [\textit{L. Denis} et al., Inverse Probl. 25, No.~11, Article ID 115017 (2009; Zbl 1191.65055)]).