1: \begin{abstract}

2: We consider the nonlinear inverse problem of learning a transition operator $\f{A}$ from {partial observations at different times}, in particular from {sparse observations} of entries of its powers $\f{A},\f{A}^2,\cdots,\f{A}^{T}$. This \emph{Spatio-Temporal Transition Operator Recovery} problem is motivated by the recent interest in learning time-varying graph signals that are driven by graph operators depending on the underlying graph topology. We address the nonlinearity of the problem by embedding it into a higher-dimensional space of suitable block-Hankel matrices, where it becomes a low-rank matrix completion problem, even if $\f{A}$ is of full rank.

3: For both a uniform and an adaptive random space-time sampling model, we quantify the recoverability of the transition operator via suitable measures of incoherence of these block-Hankel embedding matrices. For graph transition operators these measures of incoherence depend on the interplay between the dynamics and the graph topology. 

4: We develop a suitable non-convex iterative reweighted least squares (IRLS) algorithm, establish its quadratic local convergence, and show that, in optimal scenarios, no more than $\mathcal{O}(rn \log(nT))$ space-time samples are sufficient to ensure accurate recovery of a rank-$r$ operator $\f{A}$ of size $n \times n$. This establishes that {spatial samples} can be substituted by a comparable number of {space-time samples}.

5: We provide an efficient implementation of the proposed IRLS algorithm with space complexity of order $O(r n T)$ and per-iteration time complexity {linear} in $n$. Numerical experiments for transition operators based on several graph models confirm that the theoretical findings accurately track empirical phase transitions, and illustrate the applicability and scalability of the proposed algorithm.

6: \end{abstract}

7: