Gaussian neural networks improve neural updates with noise based guidance

Gaussian Neural Networks

Machine LearningArtificial Intelligence

Summary

Neural networks are computer systems that learn patterns but can sometimes make mistakes because they adjust their internal parts too freely. This paper introduces Gaussian neural networks, which add a way to gently guide these adjustments by considering how surprising the network’s activities are, treating them like noisy signals. This makes learning more stable and helps the network make better predictions, especially when it comes to understanding how uncertain those predictions are. The authors show that these networks work better on various tasks compared to traditional methods.

What this means in practice

  • For machine learning engineers: Train more reliable neural networks by incorporating Gaussian activity priors to improve performance and uncertainty estimation on classification and regression tasks.
  • For computer vision developers: Enhance image recognition systems with Gaussian neural networks that reduce overfitting and better handle surprising inputs.

Authors

Peter Kuhn, Victoria Heusinger-Heß

Abstract

Gaussian neural networks (GaNNs) are proposed as a novel regularization mechanism for neural networks. From a Bayesian perspective standard regularization techniques can be viewed as imposing priors over weight-space. Assuming priors over activation-space remains a largely unexplored possibility. GaNNs assume such priors. They do this by treating activities from earlier layers like signals with Gaussian noise and predicting the properties of the noise distribution using an additional unsupervised loss. While training, the unsupervised loss acts as a penalty on unexpected activities, allowing greater weight updates in less surprising directions. The paper demonstrates the superiority of Gaussian neural networks over standard neural networks on a variety of classification and regression tasks. We also investigate the ability of GaNNs to quantify uncertainty.