Pseudo-D: Informing Multi-View Uncertainty Estimation with Calibrated Neural Training Dynamics

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gu, Ang Nan, Tsang, Michael, Vaseli, Hooman, Abolmaesumi, Purang, Tsang, Teresa
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914038189916160
author Gu, Ang Nan
Tsang, Michael
Vaseli, Hooman
Abolmaesumi, Purang
Tsang, Teresa
author_facet Gu, Ang Nan
Tsang, Michael
Vaseli, Hooman
Abolmaesumi, Purang
Tsang, Teresa
contents Computer-aided diagnosis systems must make critical decisions from medical images that are often noisy, ambiguous, or conflicting, yet today's models are trained on overly simplistic labels that ignore diagnostic uncertainty. One-hot labels erase inter-rater variability and force models to make overconfident predictions, especially when faced with incomplete or artifact-laden inputs. We address this gap by introducing a novel framework that brings uncertainty back into the label space. Our method leverages neural network training dynamics (NNTD) to assess the inherent difficulty of each training sample. By aggregating and calibrating model predictions during training, we generate uncertainty-aware pseudo-labels that reflect the ambiguity encountered during learning. This label augmentation approach is architecture-agnostic and can be applied to any supervised learning pipeline to enhance uncertainty estimation and robustness. We validate our approach on a challenging echocardiography classification benchmark, demonstrating superior performance over specialized baselines in calibration, selective classification, and multi-view fusion.
format Preprint
id arxiv_https___arxiv_org_abs_2509_11800
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Pseudo-D: Informing Multi-View Uncertainty Estimation with Calibrated Neural Training Dynamics
Gu, Ang Nan
Tsang, Michael
Vaseli, Hooman
Abolmaesumi, Purang
Tsang, Teresa
Computer Vision and Pattern Recognition
Computer-aided diagnosis systems must make critical decisions from medical images that are often noisy, ambiguous, or conflicting, yet today's models are trained on overly simplistic labels that ignore diagnostic uncertainty. One-hot labels erase inter-rater variability and force models to make overconfident predictions, especially when faced with incomplete or artifact-laden inputs. We address this gap by introducing a novel framework that brings uncertainty back into the label space. Our method leverages neural network training dynamics (NNTD) to assess the inherent difficulty of each training sample. By aggregating and calibrating model predictions during training, we generate uncertainty-aware pseudo-labels that reflect the ambiguity encountered during learning. This label augmentation approach is architecture-agnostic and can be applied to any supervised learning pipeline to enhance uncertainty estimation and robustness. We validate our approach on a challenging echocardiography classification benchmark, demonstrating superior performance over specialized baselines in calibration, selective classification, and multi-view fusion.
title Pseudo-D: Informing Multi-View Uncertainty Estimation with Calibrated Neural Training Dynamics
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.11800