State of the Art Uncertainty

State-of-the-art uncertainty quantification in deep learning focuses on reliably estimating the confidence of model predictions, particularly in high-stakes applications where trust is paramount. Current research emphasizes methods like Bayesian neural networks, deep ensembles, and novel approaches integrating uncertainty directly into model architectures (e.g., through latent variable evolution or function-space inference). These advancements aim to improve the robustness and trustworthiness of AI systems across diverse fields, from medical diagnosis and autonomous driving to scientific modeling and air quality forecasting, by providing a principled measure of prediction uncertainty. Addressing the limitations of existing methods, such as computational cost and vulnerability to adversarial attacks, remains a key challenge.

Papers

February 13, 2024

Uncertainty Quantification for Forward and Inverse Problems of PDEs via Latent Global Evolution
Tailin Wu, Willie Neiswanger, Hongtao Zheng, Stefano Ermon, Jure Leskovec
Uncertainty Quantification Inverse Problem Partial Differential Equation Deep Learning Based Surrogate State of the Art Uncertainty Learning Based Surrogate Latent Evolution

December 28, 2023

Tractable Function-Space Variational Inference in Bayesian Neural Networks
Tim G. J. Rudner, Zonghao Chen, Yee Whye Teh, Yarin Gal
Bayesian Neural Network Predictive Uncertainty Reliable Uncertainty State of the Art Uncertainty Function Space Variational Inference

October 19, 2023

Be Bayesian by Attachments to Catch More Uncertainty
Shiyu Shen, Bin Pan, Tianyang Shi, Tao Li, Zhenwei Shi
Uncertainty Estimation Bayesian Neural Network Document Image Bayesian Structure State of the Art Uncertainty

September 29, 2023

AdaDiff: Accelerating Diffusion Models through Step-Wise Adaptive Computation
Shengkun Tang, Yaqing Wang, Caiwen Ding, Yi Liang, Yao Li, Dongkuan Xu
Diffusion Model High Resolution Image Uncertainty Aware Intermediate Layer Early eXit State of the Art Uncertainty Layer Wise Loss

August 1, 2023

Capsa: A Unified Framework for Quantifying Risk in Deep Neural Networks
Sadhana Lolla, Iaroslav Elistratov, Alejandro Perez, Elaheh Ahmadi, Daniela Rus, Alexander Amini
Deep Neural Network Large Scale Unified Framework Epistemic Uncertainty High Quality Risk Description Risk Awareness State of the Art Uncertainty

June 22, 2023

Harnessing Data Augmentation to Quantify Uncertainty in the Early Estimation of Single-Photon Source Quality
David Jacob Kedziora, Anna Musiał, Wojciech Rudno-Rudziński, Bogdan Gabrys
Data Augmentation Estimation Task Experimental Validation Single Photon State of the Art Uncertainty

October 3, 2022

On Attacking Out-Domain Uncertainty Estimation in Deep Neural Networks
Huimin Zeng, Zhenrui Yue, Yang Zhang, Ziyi Kou, Lanyu Shang, Dong Wang
Deep Neural Network Uncertainty Estimation Parameter Uncertainty Reliable Uncertainty Robust Uncertainty State of the Art Uncertainty

August 11, 2022

Region-Based Evidential Deep Learning to Quantify Uncertainty and Improve Robustness of Brain Tumor Segmentation
Hao Li, Yang Nan, Javier Del Ser, Guang Yang
Native Robustness Uncertainty Estimation Brain Tumor Segmentation Evidential Deep Learning Robust Segmentation Segmentation Uncertainty State of the Art Uncertainty

August 3, 2022

Gradient-based Uncertainty for Monocular Depth Estimation
Julia Hornauer, Vasileios Belagiannis
High Uncertainty Anticipation Monocular Depth Estimation Depth Prediction Depth Estimation Model Auxiliary Loss State of the Art Uncertainty

December 5, 2021

Probabilistic Deep Learning to Quantify Uncertainty in Air Quality Forecasting
Abdulmajid Murad, Frank Alexander Kraemer, Kerstin Bach, Gavin Taylor
Uncertainty Quantification Model Uncertainty Predictive Uncertainty Air Quality Forecasting Probabilistic Deep Learning State of the Art Uncertainty Deep Probabilistic Forecasting