Heteroscedastic Uncertainty

Heteroscedastic uncertainty modeling focuses on representing the varying levels of uncertainty present in data, acknowledging that prediction errors are not uniformly distributed. Current research emphasizes improving the accuracy and robustness of heteroscedastic uncertainty estimation within Bayesian neural networks and other probabilistic models, often employing techniques like moment propagation or adaptive weighting schemes to address challenges such as noisy gradients and undersampling. This work is crucial for enhancing the reliability of machine learning applications across diverse fields, including medical image registration, UAV navigation, and speech enhancement, by providing more informative and nuanced uncertainty quantification. The ultimate goal is to build more trustworthy and reliable AI systems that can better account for and communicate their limitations.

Papers

February 22, 2024

A Framework for Variational Inference of Lightweight Bayesian Neural Networks with Heteroscedastic Uncertainties
David J. Schodt, Ryan Brown, Michael Merritt, Samuel Park, Delsin Menolascino, Mark A. Peot
Neural Network New Framework Variational Inference Bayesian Neural Network Heteroscedastic Uncertainty

December 1, 2023

Heteroscedastic Uncertainty Estimation for Probabilistic Unsupervised Registration of Noisy Medical Images
Xiaoran Zhang, Daniel H. Pak, Shawn S. Ahn, Xiaoxiao Li, Chenyu You, Lawrence Staib, Albert J. Sinusas, Alex Wong, James S. Duncan
Medical Image Image Registration Heteroscedastic Noise Noisy MRI Heteroscedastic Uncertainty Displacement Estimation

September 25, 2023

Perception-and-Energy-aware Motion Planning for UAV using Learning-based Model under Heteroscedastic Uncertainty
Reiya Takemura, Genya Ishigami
Unmanned Aerial Vehicle Motion Planning Learning Based Trajectory Planning Heteroscedastic Uncertainty

November 16, 2022

Leveraging Heteroscedastic Uncertainty in Learning Complex Spectral Mapping for Single-channel Speech Enhancement
Kuan-Lin Chen, Daniel D. E. Wong, Ke Tan, Buye Xu, Anurag Kumar, Vamsi Krishna Ithapu
Single Channel Speech Enhancement Multivariate Gaussian Negative Log Likelihood Heteroscedastic Uncertainty

March 17, 2022

On the Pitfalls of Heteroscedastic Uncertainty Estimation with Probabilistic Neural Networks
Maximilian Seitzer, Arash Tavakoli, Dimitrije Antic, Georg Martius
Common Pitfall Aleatoric Uncertainty Log Likelihood Probabilistic Neural Network Likelihood Loss Heteroscedastic Regression Heteroscedastic Uncertainty

November 22, 2021

Dense Uncertainty Estimation via an Ensemble-based Conditional Latent Variable Model
Jing Zhang, Yuchao Dai, Mehrtash Harandi, Yiran Zhong, Nick Barnes, Richard Hartley
Epistemic Uncertainty Conditional Generative Aleatoric Uncertainty Dense Recognition Heteroscedastic Uncertainty

Heteroscedastic Uncertainty

Papers

A Framework for Variational Inference of Lightweight Bayesian Neural Networks with Heteroscedastic Uncertainties

Heteroscedastic Uncertainty Estimation for Probabilistic Unsupervised Registration of Noisy Medical Images

Perception-and-Energy-aware Motion Planning for UAV using Learning-based Model under Heteroscedastic Uncertainty

Leveraging Heteroscedastic Uncertainty in Learning Complex Spectral Mapping for Single-channel Speech Enhancement

On the Pitfalls of Heteroscedastic Uncertainty Estimation with Probabilistic Neural Networks

Dense Uncertainty Estimation via an Ensemble-based Conditional Latent Variable Model