On polyhedral estimation of signals via indirect observations (Q2293715)

The author develops a simple generic efficiently computable nonlinear in observations ``polyhedral'' estimate along with computation-friendly techniques for its design and risk analysis. Assumptions \begin{itemize} \item a nonempty computationally tractable convex compact signal set \(X \subset \mathbb R^n\), \item sensing matrix \(A \in \mathbb R^{m\times n}\), decoding matrix \(B \in \mathbb R^{\nu\times n}\), \end{itemize} \[\zeta= A_x + \xi_x \tag{1}\] Estimation problem (1) is a (nonparametric) linear inverse problem. The problem, that is to recover \(w = B_x\) for a given observation \(\zeta\), where \(x\) is the signal underlying the observation, has been solved by author.