Suppose $A$ is a matrix consisting of real numbers and nan
's. What are some of the robust formulations and the associated algorithms for estimating its singular value decomposition? What are the references?
One formulation could be the following.
Let $I:=\{(i,j)|A_{i,j}\in \mathbf R\}$, $\mathscr U:=\{U|U^TU=I\}$, $\mathscr D:=$ set of diagonal matrices and $$f(U,\Lambda,V):=\sum_{(i,j)\in I}\big(A_{i,j}-(U\Lambda V^T)_{i,j}\big)^2.$$ We want to solve $$\min_{U,V\in\mathscr U,\,\Lambda\in\mathscr D} f(U,\Lambda,V).$$