Navigating Noise: A Study of How Noise Influences Generalisation and Calibration of Neural Networks

TL;DR

This paper investigates how different noise types and placements during neural network training affect their ability to generalise and calibrate confidence, highlighting domain-specific effectiveness and the complexity of joint optimisation.

Contribution

It provides a comprehensive analysis of diverse noise modalities, their impacts on generalisation and calibration, and insights into domain-specific tuning and hyperparameter transferability.

Findings

AugMix and weak augmentation are effective across multiple computer vision tasks.

Combining noise types can improve generalisation and calibration within a domain.

Transferring noise benefits across different domains remains challenging.

Abstract

Enhancing the generalisation abilities of neural networks (NNs) through integrating noise such as MixUp or Dropout during training has emerged as a powerful and adaptable technique. Despite the proven efficacy of noise in NN training, there is no consensus regarding which noise sources, types and placements yield maximal benefits in generalisation and confidence calibration. This study thoroughly explores diverse noise modalities to evaluate their impacts on NN's generalisation and calibration under in-distribution or out-of-distribution settings, paired with experiments investigating the metric landscapes of the learnt representations across a spectrum of NN architectures, tasks, and datasets. Our study shows that AugMix and weak augmentation exhibit cross-task effectiveness in computer vision, emphasising the need to tailor noise to specific domains. Our findings emphasise the efficacy of combining noises and successful hyperparameter transfer within a single domain but the difficulties in transferring the benefits to other domains. Furthermore, the study underscores the complexity of simultaneously optimising for both generalisation and calibration, emphasising the need for practitioners to carefully consider noise combinations and hyperparameter tuning for optimal performance in specific tasks and datasets.