Weight Compander: A Simple Weight Reparameterization for Regularization

TL;DR

This paper introduces weight compander, a simple reparameterization technique for weights in neural networks that enhances regularization by controlling weight magnitudes, leading to better generalization and feature extraction.

Contribution

The paper proposes a novel weight reparameterization method called weight compander, which improves regularization by implicitly restricting weight magnitudes and can be combined with other techniques.

Findings

Improves neural network generalization performance.

Encourages more feature extraction by increasing weight redundancy.

Compatible with existing regularization methods.

Abstract

Regularization is a set of techniques that are used to improve the generalization ability of deep neural networks. In this paper, we introduce weight compander (WC), a novel effective method to improve generalization by reparameterizing each weight in deep neural networks using a nonlinear function. It is a general, intuitive, cheap and easy to implement method, which can be combined with various other regularization techniques. Large weights in deep neural networks are a sign of a more complex network that is overfitted to the training data. Moreover, regularized networks tend to have a greater range of weights around zero with fewer weights centered at zero. We introduce a weight reparameterization function which is applied to each weight and implicitly reduces overfitting by restricting the magnitude of the weights while forcing them away from zero at the same time. This leads to a more democratic decision-making in the network. Firstly, individual weights cannot have too much influence in the prediction process due to the restriction of their magnitude. Secondly, more weights are used in the prediction process, since they are forced away from zero during the training. This promotes the extraction of more features from the input data and increases the level of weight redundancy, which makes the network less sensitive to statistical differences between training and test data. We extend our method to learn the hyperparameters of the introduced weight reparameterization function. This avoids hyperparameter search and gives the network the opportunity to align the weight reparameterization with the training progress. We show experimentally that using weight compander in addition to standard regularization methods improves the performance of neural networks.