On the Overlooked Pitfalls of Weight Decay and How to Mitigate Them: A Gradient-Norm Perspective
This work addresses a previously overlooked issue with weight decay for deep neural network practitioners, aiming to improve convergence and generalization.
This paper identifies that weight decay can lead to large gradient norms at the end of training, indicating poor convergence and generalization. To address this, they propose Scheduled Weight Decay (SWD), a dynamic scheduler that adjusts weight decay strength based on gradient norms, leading to improved performance over constant weight decay with Adam.
Weight decay is a simple yet powerful regularization technique that has been very widely used in training of deep neural networks (DNNs). While weight decay has attracted much attention, previous studies fail to discover some overlooked pitfalls on large gradient norms resulted by weight decay. In this paper, we discover that, weight decay can unfortunately lead to large gradient norms at the final phase (or the terminated solution) of training, which often indicates bad convergence and poor generalization. To mitigate the gradient-norm-centered pitfalls, we present the first practical scheduler for weight decay, called the Scheduled Weight Decay (SWD) method that can dynamically adjust the weight decay strength according to the gradient norm and significantly penalize large gradient norms during training. Our experiments also support that SWD indeed mitigates large gradient norms and often significantly outperforms the conventional constant weight decay strategy for Adaptive Moment Estimation (Adam).