Towards Stability of Parameter-free Optimization
This addresses the problem of manual learning rate tuning for researchers and practitioners in machine learning, offering an incremental improvement in parameter-free optimization methods.
The paper tackles the challenge of hyperparameter tuning in adaptive gradient methods by proposing AdamG, a parameter-free optimizer that automatically adapts to diverse optimization problems without manual tuning. Empirical results show that AdamG achieves performance consistently on par with Adam using a manually tuned learning rate across various tasks.
Hyperparameter tuning, particularly the selection of an appropriate learning rate in adaptive gradient training methods, remains a challenge. To tackle this challenge, in this paper, we propose a novel parameter-free optimizer, \textsc{AdamG} (Adam with the golden step size), designed to automatically adapt to diverse optimization problems without manual tuning. The core technique underlying \textsc{AdamG} is our golden step size derived for the AdaGrad-Norm algorithm, which is expected to help AdaGrad-Norm preserve the tuning-free convergence and approximate the optimal step size in expectation w.r.t. various optimization scenarios. To better evaluate tuning-free performance, we propose a novel evaluation criterion, \textit{reliability}, to comprehensively assess the efficacy of parameter-free optimizers in addition to classical performance criteria. Empirical results demonstrate that compared with other parameter-free baselines, \textsc{AdamG} achieves superior performance, which is consistently on par with Adam using a manually tuned learning rate across various optimization tasks.