[论文解读] A Survey of Uncertainty in Deep Neural Networks

Jakob Gawlikowski, Cedrique Rovile Njieutcheu Tassi+12 more|arXiv (Cornell University)|Jul 7, 2021

Fault Detection and Control Systems被引用 242

一句话总结

对深度神经网络不确定性的来源、估计方法、校准以及现实世界挑战的综合综述，包括贝叶斯、集成和测试时增强方法。

ABSTRACT

Due to their increasing spread, confidence in neural network predictions became more and more important. However, basic neural networks do not deliver certainty estimates or suffer from over or under confidence. 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 a variety of 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. A comprehensive introduction to the most crucial sources of uncertainty is given and their separation into reducible model uncertainty and not reducible data uncertainty is presented. The modeling of these uncertainties based on deterministic neural networks, Bayesian neural networks, 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 the calibration of neural networks and give an overview of existing baselines and implementations. Different examples from the wide spectrum of challenges in different fields give an idea of the needs and challenges regarding uncertainties in practical applications. Additionally, the practical limitations of current methods 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.

研究动机与目标

• 识别并对深度神经网络中的不确定性来源与类型进行分类（数据不确定性与模型不确定性）。
• 调查并比较主要的不确定性估计方法（贝叶斯神经网络、集成、测试时数据增强、确定性模型）。
• 讨论不确定性估计的校准以及现实世界应用的实际基准。
• 突出深度神经网络不确定性量化的局限性、挑战以及未来研究的机会。

提出的方法

• 描述从数据到不确定性量化的四个步骤：数据获取、网络设计/训练、推断与预测不确定性建模。
• 区分导致不确定性的因素（I: 现实世界的变异性；II: 测量噪声；III: 模型结构误差；IV: 训练过程误差；V: 未知数据），以及它们的传播方式。
• 将预测不确定性分类为数据不确定性（本质）和模型不确定性（epistemic），并讨论贝叶斯与分布形式的表述。
• 给出四种不确定性估计范式：单一确定性网络、贝叶斯方法、集合、以及测试时增强，详细说明它们的权衡。
• 概述评估指标、校准技术，以及可用的实现与基准。

实验结果

研究问题

• RQ1 DNN 预测中的主要不确定性来源与类型是什么？
• RQ2 在 DNN 中建模与量化预测不确定性的主要方法有哪些？它们之间如何比较？
• RQ3 如何对 DNN 不确定性估计进行校准并在实践中评估其可靠性？
• RQ4 当前不确定性量化方法在实际应用中的实际挑战与局限性是什么？

主要发现

• DNN 预测中的不确定性源自数据不确定性（本质不确定性）和模型不确定性（epistemic），其原因与可降低性特征各不相同。
• 贝叶斯推断、集成、测试时数据增强，以及具有显式不确定性成分的确定性模型，是不确定性估计的主要建模范式。
• 不确定性估计的校准对可靠性至关重要，存在多种校准技术，以及常见的评估数据集和基准。
• 现场应用（医疗成像、机器人、地球观测）揭示了实际挑战，如领域偏移、域外输入，以及对安全关键决策的要求。
• 该论文提供了一个框架，将不确定性来源映射到实际估计方法，并在资源约束下强调何时偏好使用每种方法。