In the section Model Training: “Our deep learning model was trained using code-free deep learning (CFDL) with the Google Cloud AutoML platform ... the CFDL pla

In the section Model Training: “Our deep learning model was trained using code-free deep learning (CFDL) with the Google Cloud AutoML platform ... the CFDL platform provides the option of image upload via shell-scripting utilizing a .csv spreadsheet containing labels ... Automated machine learning was then employed, which entails neural architecture search and hyperparameter tuning.” Earlier, in the Limitations section: “The design of the CFDL model was inherently opaque due to the framework’s automated nature with respect to model architecture and hyperparameters. While this opacity is not unique to CFDL, there is potential to further reduce ML explainability due to lack of insight of model architectures and parameters employed.” Maybe there’s a whitepaper somewhere on how Google’s AutoML works?