On Distributed Online Classification in the Midst of Concept Drifts
This work addresses distributed learning under concept drift, which is incremental as it builds on existing methods for online and distributed settings.
The paper tackled the problem of distributed online classification in non-stationary environments by analyzing generalization bounds and comparing diffusion strategies to non-cooperative methods, with results illustrated through simulations.
In this work, we analyze the generalization ability of distributed online learning algorithms under stationary and non-stationary environments. We derive bounds for the excess-risk attained by each node in a connected network of learners and study the performance advantage that diffusion strategies have over individual non-cooperative processing. We conduct extensive simulations to illustrate the results.