Stochastic optimization methods such as mirror descent have wide applications due to low computational cost. Those methods have been well studied under assumption of the independent and identical distribution, and usually achieve sublinear rate of convergence. However, this assumption may be too strong and unpractical in real application scenarios. Recent researches investigate stochastic gradient descent when instances are sampled from a Markov chain. Unfortunately, few results are known for stochastic mirror descent. In the paper, we propose a new version of stochastic mirror descent termed by MarchOn in the scenario of the federated learning. Given a distributed network, the model iteratively travels from a node to one of its neighbours randomly. Furthermore, we propose a new framework to analyze MarchOn, which yields best rates of convergence for convex, strongly convex, and non-convex loss. Finally, we conduct empirical studies to evaluate the convergence of MarchOn, and validate theoretical results.

Government-funded artificial intelligence programs could soon be organized under a new effort by the General Services Administration. GSA earlier this month created the Data Federation, a site that intends to coordinate the disparate existing data-related efforts at various agencies by sharing standards, case studies and reusable tech tools. On Monday, GSA plans to announce a new community of practice, or subsection dedicated to artificial intelligence, according to Technology Transformation Service data portfolio lead Philip Ashlock. Ashlock was speaking at a Digital Government Institute conference on Thursday. The Data Federation is still in the very early stages, he said--long term, it's working with 18F and the Presidential Innovation Fellows program to develop a "maturity model" to understand how data projects tend to evolve.