Conceptual

Variability-Aware Machine Learning Model Selection

An approach that treats the choice of ML algorithm as a configurable process rather than an ad hoc decision, by modeling the commonalities and variation points of model selection - contextual factors such as data characteristics, number of features, prediction type, and their interdependencies - as a software-product-line feature model. Students learn how making these variation points explicit yields a more transparent, interpretable, and automatable basis for selecting a model, illustrated by instantiating the Scikit-Learn estimator-selection heuristics.