Hyperparameter tuning

Classification of cardiac cohorts based on morphological and hemodynamic features derived from 4D PC-MRI data

We investigate the potential of morphological and hemodynamic features extracted from measured blood flow data in the aorta to classify hear-thealthy volunteers (HHV) and patients with bicuspid aortic valve (BAV). Furthermore, we determine features that distinguish male vs. female patients and elderly HHV vs. BAV patients. We propose a data analysis pipeline for cardiac status classification, encompassing feature selection, model training, and hyperparameter tuning.

Comparing hyperparameter tuning strategies with tidymodels

We compare 5 hyperparameter tuning search strategies in terms of (a) model quality and (b) run time on 10 learning problems with 3 machine learning algorithms using the tidymodels framework. Bayesian search gave best results, while the racing methods had lowest running time. When interpreting the results, it must be taken into account that the search strategies have their own hyperparameters, which can substantially influence the results.