Data leakage inflates prediction performance in connectome-based machine learning models
Abstract Predictive modeling is a central technique in neuroimaging to identify brain-behavior relationships and test their generalizability to unseen data. However, data leakage undermines the validity of predictive models by breaching the separation between training and test data. Leakage is alway...
Uloženo v:
| Hlavní autoři: | , , , , |
|---|---|
| Médium: | Artigo |
| Jazyk: | Inglês |
| Vydáno: |
Nature Portfolio
2024-02-01
|
| Edice: | Nature Communications |
| On-line přístup: | https://doi.org/10.1038/s41467-024-46150-w |
| Tagy: |
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|