Zobrazeno 1 - 3
of 3
pro vyhledávání: '"Wenderoth, Laura"'
Autor:
Wenderoth, Laura
This study explores the potential of using training dynamics as an automated alternative to human annotation for evaluating the quality of training data. The framework used is Data Maps, which classifies data points into categories such as easy-to-le
Externí odkaz:
http://arxiv.org/abs/2411.01653
Exploring Multi-Modality Dynamics: Insights and Challenges in Multimodal Fusion for Biomedical Tasks
Autor:
Wenderoth, Laura
This paper investigates the MM dynamics approach proposed by Han et al. (2022) for multi-modal fusion in biomedical classification tasks. The MM dynamics algorithm integrates feature-level and modality-level informativeness to dynamically fuse modali
Externí odkaz:
http://arxiv.org/abs/2411.00725
Publikováno v:
Bioengineering (Basel). 2023;10(8):895
Is self-supervised deep learning (DL) for medical image analysis already a serious alternative to the de facto standard of end-to-end trained supervised DL? We tackle this question for medical image classification, with a particular focus on one of t
Externí odkaz:
http://arxiv.org/abs/2304.05163