A Concentration Bound for Distributed Stochastic Approximation
This work provides theoretical guarantees for distributed optimization algorithms, but it is incremental as it builds on classical models without major breakthroughs.
The paper analyzes a distributed stochastic approximation scheme with consensus, deriving a high probability bound for the tracking error between interpolated iterates and a limiting differential equation.
We revisit the classical model of Tsitsiklis, Bertsekas and Athans for distributed stochastic approximation with consensus. The main result is an analysis of this scheme using the ODE approach to stochastic approximation, leading to a high probability bound for the tracking error between suitably interpolated iterates and the limiting differential equation. Several future directions will also be highlighted.