Learning an Adaptive Learning Rate Schedule
This addresses the challenge of optimizing learning rates for high-dimensional, non-convex problems in machine learning, offering a generalizable solution that can transfer across datasets, though it is incremental as it builds on existing reinforcement learning methods.
The paper tackles the problem of inflexible hand-designed learning rate schedules in neural network training by proposing a reinforcement learning framework that automatically learns adaptive schedules based on training dynamics, achieving better test results on Fashion MNIST and CIFAR10 datasets.
The learning rate is one of the most important hyper-parameters for model training and generalization. However, current hand-designed parametric learning rate schedules offer limited flexibility and the predefined schedule may not match the training dynamics of high dimensional and non-convex optimization problems. In this paper, we propose a reinforcement learning based framework that can automatically learn an adaptive learning rate schedule by leveraging the information from past training histories. The learning rate dynamically changes based on the current training dynamics. To validate this framework, we conduct experiments with different neural network architectures on the Fashion MINIST and CIFAR10 datasets. Experimental results show that the auto-learned learning rate controller can achieve better test results. In addition, the trained controller network is generalizable -- able to be trained on one data set and transferred to new problems.