We introduce the Projected Push-Pull algorithm that enables multiple agents to solve a distributed constrained optimization problem with private cost functions and global constraints, in a collaborative manner. Our algorithm employs projected gradient descent to deal with constraints and a lazy update rule to control the trade-off between the consensus and optimization steps in the protocol. We prove that our algorithm achieves geometric convergence over time-varying directed graphs while ensuring that the decision variable always stays within the constraint set. We derive explicit bounds for step sizes that guarantee geometric convergence based on the strong-convexity and smoothness of cost functions, and graph properties. Moreover, we provide additional theoretical results on the usefulness of lazy updates, revealing the challenges in the analysis of any gradient tracking method that uses projection operators in a distributed constrained optimization setting. We validate our theoretical results with numerical studies over different graph types, showing that our algorithm achieves geometric convergence empirically.

We consider decentralized optimization problems in which a number of agents collaborate to minimize the average of their local functions by exchanging over an underlying communication graph. Specifically, we place ourselves in an asynchronous model where only a random portion of nodes perform computation at each iteration, while the information exchange can be conducted between all the nodes and in an asymmetric fashion. For this setting, we propose an algorithm that combines gradient tracking with a network-level variance reduction (in contrast to variance reduction within each node). This enables each node to track the average of the gradients of the objective functions. Our theoretical analysis shows that the algorithm converges linearly, when the local objective functions are strongly convex, under mild connectivity conditions on the expected mixing matrices. In particular, our result does not require the mixing matrices to be doubly stochastic. In the experiments, we investigate a broadcast mechanism that transmits information from computing nodes to their neighbors, and confirm the linear convergence of our method on both synthetic and real-world datasets.