Logit Attenuating Weight Normalization
This addresses the issue of poor generalization in deep learning for practitioners, offering an incremental improvement to optimizers.
The paper tackles the problem of over-parameterized deep networks losing adaptivity and generalizing poorly due to large logits and weights, proposing Logit Attenuating Weight Normalization (LAWN) to control logits by constraining weight norms, resulting in improved generalization performance on large-scale image classification and recommender systems, with notable gains for Adam and large batch sizes.
Over-parameterized deep networks trained using gradient-based optimizers are a popular choice for solving classification and ranking problems. Without appropriately tuned $\ell_2$ regularization or weight decay, such networks have the tendency to make output scores (logits) and network weights large, causing training loss to become too small and the network to lose its adaptivity (ability to move around) in the parameter space. Although regularization is typically understood from an overfitting perspective, we highlight its role in making the network more adaptive and enabling it to escape more easily from weights that generalize poorly. To provide such a capability, we propose a method called Logit Attenuating Weight Normalization (LAWN), that can be stacked onto any gradient-based optimizer. LAWN controls the logits by constraining the weight norms of layers in the final homogeneous sub-network. Empirically, we show that the resulting LAWN variant of the optimizer makes a deep network more adaptive to finding minimas with superior generalization performance on large-scale image classification and recommender systems. While LAWN is particularly impressive in improving Adam, it greatly improves all optimizers when used with large batch sizes