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.

An Enhanced Particle Filter for Uncertainty Quantification in Neural Networks

Estimation with particle filter under model uncertainty

A Kernel Framework to Quantify a Model's Local Predictive Uncertainty under Data Distributional Shifts

Deep Particle Filter for Vehicle Trajectory Estimation under Uncertainty

Uncertainty Quantification for Deep Learning

Density Uncertainty Layers for Reliable Uncertainty Estimation

A Survey of Uncertainty in Deep Neural Networks

Lightweight Model Uncertainty Estimation for Deep Neural Object Detection

High Accuracy Uncertainty-Aware Interatomic Force Modeling with Equivariant Bayesian Neural Networks

Differentiable and Stable Long-Range Tracking of Multiple Posterior Modes

Adaptive Particle Filter For Nonparametric Estimation With Measurement Uncertainty In Wireless Sensor Networks

Uncertainty Aware Deep Learning for Particle Accelerators

Density regression and uncertainty quantification with Bayesian deep noise neural networks

Efficient Uncertainty Quantification and Reduction for Over-Parameterized Neural Networks

Uncertainty quantification in scientific machine learning: Methods, metrics, and comparisons

A principled distance-aware uncertainty quantification approach for enhancing the reliability of physics-informed neural network

One step closer to unbiased aleatoric uncertainty estimation

Robust Particle Filter by Dynamic Averaging of Multiple Noise Models

Integrating uncertainty quantification into randomized smoothing based robustness guarantees

Approaching Neural Network Uncertainty Realism

Neural State-Space Models: Empirical Evaluation of Uncertainty Quantification