A review of uncertainty quantification in deep learning: Techniques, applications and challenges
Обзор количественной оценки неопределенности в глубоком обучении: методы, приложения и проблемы
2021-05-23
SCID: 54.1/xpq8u5mf
Discuss with AI
Bayesian approximationdeep learningensemble learningreinforcement learninguncertainty quantification
Figures from the paper
Abstract (AI)
Uncertainty quantification (UQ) methods play a pivotal role in reducing the impact of uncertainties during both optimization and decision making processes. They have been applied to solve a variety of real-world problems in science and engineering. Bayesian approximation and ensemble learning techniques are two widely-used types of uncertainty quantification (UQ) methods. In this regard, researchers have proposed different UQ methods and examined their performance in a variety of applications such as computer vision (e.g., self-driving cars and object detection), image processing (e.g., image restoration), medical image analysis (e.g., medical image classification and segmentation), natural language processing (e.g., text classification, social media texts and recidivism risk-scoring), bioinformatics, etc. This study reviews recent advances in UQ methods used in deep learning, investigates the application of these methods in reinforcement learning, and highlights fundamental research challenges and directions associated with UQ.
Key Findings
1
Bayesian approximation and ensemble learning are identified as two widely used families of uncertainty quantification methods.
2
Deep-learning UQ techniques have been applied across computer vision, image processing, medical image analysis, natural language processing, and bioinformatics.
3
Fundamental research challenges and future directions for uncertainty quantification in deep learning are highlighted.
4
The review examines how uncertainty quantification methods are used in reinforcement learning.
5
Uncertainty quantification methods reduce the impact of uncertainty in deep-learning optimization and decision-making processes.
Research Object
Uncertainty quantification (UQ) methods in deep learning
Research Subject
recent advances, applications, performance, and research challenges of uncertainty quantification methods, including their use in reinforcement learning
Publication Details
Publication Date
2021-05-23
Journal
Publisher
ISSN
Cited by
2777
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai10
Exploiting Generative AI to Scale up Intelligent Tutoring Systems2023
A unified formulation of the constant temperature molecular dynamics methods1984
Intriguing properties of neural networks2013
Gaussian Processes for Machine Learning2005
Proceedings of the 24th international conference on Machine learning2007
ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks2020
Graph neural networks: A review of methods and applications2020
Graph Neural Networks: A Review of Methods and Applications2018
Deep echo state networks with uncertainty quantification for spatio‐temporal forecasting2018
A Multi-Modal Distributed Real-Time IoT System for Urban Traffic Control (Invited Paper)2024
Cited by8
A survey of uncertainty in deep neural networks2023
Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation2023
Deep learning in optical metrology: a review2022
Bias in medical AI: Implications for clinical decision-making2024
Comparative Analysis of Intrusion Detection Systems and Machine Learning-Based Model Analysis Through Decision Tree2023
A review of Earth Artificial Intelligence2022
Integrating permutation feature importance with conformal prediction for robust Explainable Artificial Intelligence in predictive process monitoring2025
Advances in artificial intelligence for climate science methods applications and evaluation challenges2026