There has been recent progress on inferring the structure of interactions in complex networks when they are in stationary states satisfying detailed balance, but little has been done for nonequilibrium systems. Here we introduce an approach to this problem, considering, as an example, the question of recovering the interactions in an asymmetrically-coupled, synchronously-updated Sherrington-Kirkpatrick model. We derive an exact iterative inversion algorithm and develop efficient approximations based on dynamical mean-field and Thouless-Anderson-Palmer equations that express the interactions in terms of equal-time and one time step-delayed correlation functions. Introduction.-Finding the connectivity in complex networks is crucial for understanding how they operate. Gene and multi-electrode microarrays have recently made the type of data required for this purpose available. What is needed now is appropriate theoretical tools for analyzing these data and extracting the connectivity.In much recent work on this subject [1-3], the problem has been posed as that of inferring the parameters of a stationary Gibbs distribution modeling the system. While satisfied in many applications, the assumption of Gibbs equilibrium is unlikely to hold for many biological systems since they are usually driven by time-dependent external fields, their interactions may not satisfy detailed balance, or they may only be observed while the transients dominate the dynamics. Applying the equilibrium approach to such cases usually yields effective interactions that do not bear an obvious relationship to the real ones [3]. Kinetic and nonequilibrium models provide a much richer platform for studying such systems [4][5][6].Whereas for equilibrium models the development of systematic mean field inference methods [7] has led to great practical and conceptual advancements, a mean field theory for nonequilibrium network reconstruction is still lacking. In this paper, we show how a mean field theory for inference can also be developed for a nonequilibrium system. We consider this problem for a particular simple nonequilibrium model: a kinetic Ising model with random asymmetric interactions (J ji independent of J ij ), in an external field which may be time-dependent. This is a discrete-time, synchronously updated model composed of N spins s i = ±1 with transition probability