An Asymptotically Optimal Multi-Armed Bandit Algorithm and Hyperparameter Optimization
This addresses the critical model selection problem in deep learning for researchers and practitioners, offering an efficient solution for hyperparameter optimization.
The paper tackles the problem of hyperparameter search in deep learning by proposing a bandit-based algorithm called Sub-Sampling (SS) that is theoretically optimal in cumulative regret, and combines it with Bayesian Optimization into BOSS, which shows superior performance in applications like Neural Architecture Search and Object Detection.
The evaluation of hyperparameters, neural architectures, or data augmentation policies becomes a critical model selection problem in advanced deep learning with a large hyperparameter search space. In this paper, we propose an efficient and robust bandit-based algorithm called Sub-Sampling (SS) in the scenario of hyperparameter search evaluation. It evaluates the potential of hyperparameters by the sub-samples of observations and is theoretically proved to be optimal under the criterion of cumulative regret. We further combine SS with Bayesian Optimization and develop a novel hyperparameter optimization algorithm called BOSS. Empirical studies validate our theoretical arguments of SS and demonstrate the superior performance of BOSS on a number of applications, including Neural Architecture Search (NAS), Data Augmentation (DA), Object Detection (OD), and Reinforcement Learning (RL).