Total Uncertainty

Total uncertainty quantification aims to comprehensively assess all sources of error in predictions made by computational models, encompassing inherent randomness (aleatoric uncertainty), knowledge limitations (epistemic uncertainty), and model inaccuracies (model-form uncertainty). Current research focuses on decomposing and managing these uncertainty types within various model architectures, including Bayesian neural networks, physics-informed approaches like Monte Carlo methods, and deep learning techniques such as masked image modeling. This work is crucial for improving the reliability and trustworthiness of AI systems across diverse fields, from autonomous driving and medical diagnosis to scientific simulations, by providing a more complete picture of prediction confidence and limitations.

Papers

August 13, 2024

Quantification of total uncertainty in the physics-informed reconstruction of CVSim-6 physiology
Mario De Florio, Zongren Zou, Daniele E. Schiavazzi, George Em Karniadakis
Uncertainty Quantification Model Uncertainty Physic Based Regularization Total Uncertainty

August 1, 2024

Generalisation of Total Uncertainty in AI: A Theoretical Study
Keivan Shariatmadar
Artificial Intelligence Strong Generalization High Uncertainty Anticipation Theoretical Analysis Intrinsic Uncertainty Latent Imagination Total Uncertainty

February 27, 2024

Masked Gamma-SSL: Learning Uncertainty Estimation via Masked Image Modeling
David S. W. Williams, Matthew Gadd, Paul Newman, Daniele De Martini
Masked Image Modeling Semantic Segmentation Network Segmentation Benchmark Road Driving Total Uncertainty

May 24, 2023

Neural network reconstruction of cosmology using the Pantheon compilation
Konstantinos F. Dialektopoulos, Purba Mukherjee, Jackson Levi Said, Jurgen Mifsud
Neural Network Astronomical Data Planning Based Compilation Total Uncertainty

April 14, 2022

Methodical Advice Collection and Reuse in Deep Reinforcement Learning
Sahir, Ercüment İlhan, Srijita Das, Matthew E. Taylor
Reinforcement Learning Deep Reinforcement Learning Sample Efficiency Reusable Data Exhaust Deep Reinforcement Learning Agent Suggestion System Total Uncertainty Broad Persistent Advice

March 2, 2022

MUAD: Multiple Uncertainties for Autonomous Driving, a benchmark for multiple uncertainty types and tasks
Gianni Franchi, Xuanlong Yu, Andrei Bursuc, Angel Tena, Rémi Kazmierczak, Séverine Dubuisson, Emanuel Aldea, David Filliat
Autonomous Driving New Benchmark MAESTRO Dataset New Task Predictive Uncertainty Uncertainty Decomposition Total Uncertainty

February 25, 2022

Bayesian autoencoders with uncertainty quantification: Towards trustworthy anomaly detection
Bang Xiang Yong, Alexandra Brintrup
Uncertainty Quantification Unsupervised Anomaly Detection Trustworthy Machine Learning Deep Autoencoders Total Uncertainty