A plausible representation of the relational information among entities in dynamic systems such as a living cell or a social community is a stochastic network that is topologically rewiring and semantically evolving over time. Although there is a rich literature in modeling static or temporally invariant networks, little has been done toward recovering the network structure when the networks are not observable in a dynamic context. In this article, we present a machine learning method called TESLA, which builds on a temporally smoothed l1-regularized logistic regression formalism that can be cast as a standard convex-optimization problem and solved efficiently by using generic solvers scalable to large networks. We report promising results on recovering simulated timevarying networks and on reverse engineering the latent sequence of temporally rewiring political and academic social networks from longitudinal data, and the evolving gene networks over >4,000 genes during the life cycle of Drosophila melanogaster from a microarray time course at a resolution limited only by sample frequency.evolving network ͉ social network ͉ gene network ͉ lasso ͉ Markov random field I n many problems arising in social, biological, and other fields, it is often necessary to analyze populations of entities (e.g., individuals, genes) interconnected by a set of relationships (e.g., friendship, communication, influence) represented as a network. Real-time analysis of network data is important for detecting anomalies, predicting vulnerability, and assessing the potential impact of interventions in various social, biological, or engineering systems. It is not unusual for network data to be large, dynamic, heterogeneous, noisy, and incomplete. Each of these characteristics adds a degree of complexity to the interpretation and analysis of networks.Classical network analyses mostly assume that the networks in question are fully observable, static, and isotropic. Given the heterogeneity and complexity of network data in many domains, these assumptions are limiting. For example, a majority of current investigations of biological regulatory circuitry focus on networks with invariant topology over a given set of molecules, such as a protein-protein interaction network over all proteins of an organism regardless of the conditions under which individual interactions may take place or a static gene network inferred from microarray data even though the samples may be collected over a time course or multiple conditions. In reality, over the course of a cellular process, such as a cell cycle or an immune response, there may exist multiple underlying ''themes'' that determine the functionalities of each molecule and their relationships to each other, and such themes are dynamic and stochastic. As a result, the molecular networks at each time point are context dependent and can undergo systematic rewiring, rather than being invariant over time. Indeed, in a seminal study by Luscombe et al. (1), it was shown that the ''active regulatory paths'' in a ge...