Lazy Lagrangians with Predictions for Online Learning
This work addresses constrained online learning problems for applications like resource allocation, offering an incremental improvement by extending the FTRL framework with predictions.
The paper tackles online convex optimization with time-varying constraints and predictions, proposing a primal-dual algorithm that achieves tunable regret and constraint violation bounds, with constant factors shrinking with prediction quality, eventually reaching O(1) regret for perfect predictions.
We consider the general problem of online convex optimization with time-varying additive constraints in the presence of predictions for the next cost and constraint functions. A novel primal-dual algorithm is designed by combining a Follow-The-Regularized-Leader iteration with prediction-adaptive dynamic steps. The algorithm achieves $\mathcal O(T^{\frac{3-β}{4}})$ regret and $\mathcal O(T^{\frac{1+β}{2}})$ constraint violation bounds that are tunable via parameter $β\!\in\![1/2,1)$ and have constant factors that shrink with the predictions quality, achieving eventually $\mathcal O(1)$ regret for perfect predictions. Our work extends the FTRL framework for this constrained OCO setting and outperforms the respective state-of-the-art greedy-based solutions, without imposing conditions on the quality of predictions, the cost functions or the geometry of constraints, beyond convexity.