Invasive or More Direct Measurements Can Provide an Objective Early-Stopping Ceiling for Training Deep Neural Networks on Non-invasive or Less-Direct Biomedical Data

Early stopping is an extremely common tool to minimize overfitting, which would otherwise be a cause of poor generalization of the model to novel data. However, early stopping is a heuristic that, while effective, primarily relies on ad hoc parameters and metrics. Optimizing when to stop remains a c...

Descripción completa

Detalles Bibliográficos
Autores principales: Bartlett, Christopher W., Bossenbroek, Jamie, Ueyama, Yukie, McCallinhart, Patricia, Peters, Olivia A., Santillan, Donna A., Santillan, Mark K., Trask, Aaron J., Ray, William C.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Springer Nature Singapore 2023
Materias:
Original Research
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9836982/
https://www.ncbi.nlm.nih.gov/pubmed/36647373
http://dx.doi.org/10.1007/s42979-022-01553-8

Internet

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9836982/
https://www.ncbi.nlm.nih.gov/pubmed/36647373
http://dx.doi.org/10.1007/s42979-022-01553-8

Ejemplares similares