Ensemble Uncertainty

Ensemble uncertainty quantifies the reliability of predictions from multiple deep learning models, aiming to improve the robustness and safety of AI systems. Current research focuses on leveraging ensemble methods, including Monte Carlo Dropout and variations employing diverse priors or random activation functions, to estimate uncertainty in diverse applications like reinforcement learning, image processing, and physical system modeling. This work is crucial for building trust in AI systems by providing a measure of confidence in their predictions, leading to safer and more reliable applications across various fields.

Papers

October 27, 2024

Evaluation of uncertainty estimations for Gaussian process regression based machine learning interatomic potentials
Matthias Holzenkamp, Dongyu Lyu, Ulrich Kleinekathöfer, Peter Zaspel
Global Evaluation New Machine Uncertainty Estimation Gaussian Process Regression Interatomic Potential Uncertain Sample Ensemble Uncertainty

March 9, 2024

Providing Safety Assurances for Systems with Unknown Dynamics
Hao Wang, Javier Borquez, Somil Bansal
System Description Autonomous System Model Uncertainty Unknown Dynamic Learned Dynamic Safety Assurance Ensemble Uncertainty

October 23, 2023

Diverse Priors for Deep Reinforcement Learning
Chenfan Weng, Zhongguo Li
Reinforcement Learning Deep Reinforcement Learning Model Based Prior Ensemble Based Ensemble Uncertainty

August 20, 2023

Homogenising SoHO/EIT and SDO/AIA 171\AA$~$ Images: A Deep Learning Approach
Subhamoy Chatterjee, Andrés Muñoz-Jaramillo, Maher Dayeh, Hazel M. Bain, Kimberly Moreland
Deep Learning Approach Model Ensemble Electrical Impedance Tomography Solar Dynamic Observatory Space Weather Ensemble Uncertainty

April 4, 2023

Uncertainty estimation in Deep Learning for Panoptic segmentation
Michael Smith, Frank Ferrie
Deep Learning Uncertainty Estimation Panoptic Segmentation Ensemble Based Deep Learning Based Computer Vision Ensemble Uncertainty

February 28, 2023

Toward Robust Uncertainty Estimation with Random Activation Functions
Yana Stoyanova, Soroush Ghandi, Maryam Tavakol
Uncertainty Quantification Ensemble Model Ensemble Diversity Robust Uncertainty Ensemble Uncertainty

February 20, 2023

Quantifying uncertainty for deep learning based forecasting and flow-reconstruction using neural architecture search ensembles
Romit Maulik, Romain Egele, Krishnan Raghavan, Prasanna Balaprakash
Deep Learning Uncertainty Quantification Neural Architecture Search High Uncertainty Anticipation State of the Art Forecasting Ensemble Based Neural Network Ensemble Novel Ensemble Flow Reconstruction Ensemble Uncertainty

August 19, 2022

Ensemble uncertainty as a criterion for dataset expansion in distinct bone segmentation from upper-body CT images
Eva Schnider, Antal Huck, Mireille Toranelli, Georg Rauter, Azhar Zam, Magdalena Müller-Gerbl, Philippe Cattin
CT Image Bone Tissue Bone Segmentation Front Door Criterion Ensemble Uncertainty