Predictor-corrector algorithms for stochastic optimization under gradual distribution shift
This addresses optimization challenges in machine learning applications like domain shift and tracking, but appears incremental as it builds on existing predictor-corrector concepts.
The paper tackles the problem of time-varying stochastic optimization under gradual distribution shift by developing predictor-corrector algorithms that exploit underlying continuity, showing theoretically and empirically that these methods outperform non-predictor corrector approaches.
Time-varying stochastic optimization problems frequently arise in machine learning practice (e.g. gradual domain shift, object tracking, strategic classification). Although most problems are solved in discrete time, the underlying process is often continuous in nature. We exploit this underlying continuity by developing predictor-corrector algorithms for time-varying stochastic optimizations. We provide error bounds for the iterates, both in presence of pure and noisy access to the queries from the relevant derivatives of the loss function. Furthermore, we show (theoretically and empirically in several examples) that our method outperforms non-predictor corrector methods that do not exploit the underlying continuous process.