Artificial intelligence faces reproducibility crisis.
By M. Hutson. Science, 2018
600 citations recorded by the source corpus.
Subfields: Health Informatics
View Paper: https://www.semanticscholar.org/paper/770058895898d098f149dafc86af9caff92f8427
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Health Informatics
Citations in the corpus
- The AI gambit: leveraging artificial intelligence to combat climate change—opportunities, challenges, and recommendations (2021, background): “…to hyperparameters, or of the random numbers generated to start the training process in the case of RL, is essential to allow researchers to reproduce results without going through a long, and environmentally costly, process of trial and error (Hutson 2018; Strubell et al. 2019; Gibney 2020).”
- Machine Learning in Drug Discovery: A Review (2021, neutral): “2016), model reproducib0ility (Hutson 2018).”
- Convolutional Neural Networks for the Automatic Identification of Plant Diseases (2019, background): “To ensure the results obtained and shared are useful to the scientific community, it is important to have a reproducible research perspective, which is sometimes still lacking in artificial intelligence (Hutson, 2018).”
- Deep Learning in Alzheimer's Disease: Diagnostic Classification and Prognostic Prediction Using Neuroimaging Data (2019, background): “Deep learning performance is sensitive to the random numbers generated at the start of training, and hyper-parameters, such as learning rates, batch sizes, weight decay, momentum, and dropout probabilities, may be tuned by practitioners (Hutson, 2018).”
- Reporting Guidelines to Increase the Reproducibility and Comparability of Research on Microplastics (2020, background): “Issues with reproducibility and comparability of both data and methods are common across all scientific fields, 2–4 including microplastic research.”