Weight Clipping for Deep Continual and Reinforcement Learning
This addresses overfitting and plasticity issues in deep learning systems, offering a low-complexity solution for practitioners, though it is incremental as it builds on existing methods.
The paper tackles learning failures in deep continual and reinforcement learning caused by increasing weight magnitudes, proposing weight clipping as a simple add-on technique. Empirical results show benefits for generalization, addressing loss of plasticity and policy collapse, and enabling learning with a large replay ratio.
Many failures in deep continual and reinforcement learning are associated with increasing magnitudes of the weights, making them hard to change and potentially causing overfitting. While many methods address these learning failures, they often change the optimizer or the architecture, a complexity that hinders widespread adoption in various systems. In this paper, we focus on learning failures that are associated with increasing weight norm and we propose a simple technique that can be easily added on top of existing learning systems: clipping neural network weights to limit them to a specific range. We study the effectiveness of weight clipping in a series of supervised and reinforcement learning experiments. Our empirical results highlight the benefits of weight clipping for generalization, addressing loss of plasticity and policy collapse, and facilitating learning with a large replay ratio.