A survey of uncertainty in deep neural networks

Abstract

Over the last decade, neural networks have reached almost every field of science and become a crucial part of various real world applications. Due to the increasing spread, confidence in neural network predictions has become more and more important. However, basic neural networks do not deliver certainty estimates or suffer from over- or under-confidence, i.e. are badly calibrated. To overcome this, many researchers have been working on understanding and quantifying uncertainty in a neural network’s prediction. As a result, different types and sources of uncertainty have been identified and various approaches to measure and quantify uncertainty in neural networks have been proposed. This work gives a comprehensive overview of uncertainty estimation in neural networks, reviews recent advances in the field, highlights current challenges, and identifies potential research opportunities. It is intended to give anyone interested in uncertainty estimation in neural networks a broad overview and introduction, without presupposing prior knowledge in this field. For that, a comprehensive introduction to the most crucial sources of uncertainty is given and their separation into reducible model uncertainty and irreducible data uncertainty is presented. The modeling of these uncertainties based on deterministic neural networks, Bayesian neural networks (BNNs), ensemble of neural networks, and test-time data augmentation approaches is introduced and different branches of these fields as well as the latest developments are discussed. For a practical application, we discuss different measures of uncertainty, approaches for calibrating neural networks, and give an overview of existing baselines and available implementations. Different examples from the wide spectrum of challenges in the fields of medical image analysis, robotics, and earth observation give an idea of the needs and challenges regarding uncertainties in the practical applications of neural networks. Additionally, the practical limitations of uncertainty quantification methods in neural networks for mission- and safety-critical real world applications are discussed and an outlook on the next steps towards a broader usage of such methods is given.

A Deep Neural Network – Mechanistic HybridModel to Predict Pharmacokinetics in Rat

“…But, approaches with statistical correct Bayesian or Frequentist epistemic uncertainty provide better uncertainty estimates. We expect that in both cases well calibrated uncertainties can be provided by computationally expensive ensembles techniques [37][38][39].…”

Data‐driven predictive modeling in risk assessment: Challenges and directions for proper uncertainty representation

“…Therefore, in the following, we only consider noninformative prior distributions, such as uniform distributions, which typically are used when there is no prior information about the parameters. By applying the Bayes theorem and given training data, D = (x, y), the prior distribution can be updated to a posterior probability distribution p(θ|D) = p(θ|x, y) (Gawlikowski et al, 2021). Then, given new observations of the input data, x', a probability distribution over a new realization, y', of the model output is given by:…”

An Ensemble of Neural Networks for Moist Physics Processes, Its Generalizability and Stable Integration

“…Brenowitz and Bretherton (2019) first noted that the training bias fluctuates significantly from one training epoch to another, and thus determining when to stop the training can lead to considerable uncertainties. Furthermore, individual predictions from a deep learning model can contain sizable uncertainties even though the model performs well on average (Gawlikowski et al., 2021; Pearce et al., 2018). The prediction uncertainty from NN‐based parameterizations can come from two major sources: aleatoric and epistemic uncertainties.…”