Checklist Strategies to Improve the Reproducibility of Deep Learning Experiments with an Illustration
Résumé
We report a review of the reproducibility of three publications for Poverty estimation using DL and Remote sensing imagery. For each experiment, we identified the methods and workflows used, if the experiments were not fully reproducible. Although the three use cases were proposed for a specific task (poverty estimation), we believe that the evaluation methods could be applied to more general Deep Learning tasks, where difficulties might include (a) a lack of dataset specificity (and the metadata related with it), (b) inadequate description of the DL methodology, (c) the implementation methodology, and the infrastructure used. We also feel that these recommendations can be extended to other domains such as medical, climatic, biodiversity, industrial, military, etc.
Fichier principal
Checklist Strategies to Improve the Reproducibility of Deep Learning Experiments with an illustration (1).pdf (1.13 Mo)
Télécharger le fichier
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|
