Tracking Performance of Online Stochastic Learners
This work provides a theoretical framework for analyzing tracking performance in online learning, which is incremental as it builds on existing adaptive filter analogies.
The paper tackled the problem of tracking optimal solutions in online stochastic learners under non-stationary conditions, establishing a link between steady-state performance and tracking performance that allows for direct inference from steady-state expressions.
The utilization of online stochastic algorithms is popular in large-scale learning settings due to their ability to compute updates on the fly, without the need to store and process data in large batches. When a constant step-size is used, these algorithms also have the ability to adapt to drifts in problem parameters, such as data or model properties, and track the optimal solution with reasonable accuracy. Building on analogies with the study of adaptive filters, we establish a link between steady-state performance derived under stationarity assumptions and the tracking performance of online learners under random walk models. The link allows us to infer the tracking performance from steady-state expressions directly and almost by inspection.