Introduction¶
Model search result and reporting¶
The class GridSearchCV
of scikit-learn provides
an easy way to get the cross-validation scores of an
estimator for a grid of hyper-parameters. While this is
a usefull feature, an often scenario in machine learning
research is to design experiments where multiple estimators
are compared for various hyper-parameter grids. This functionality
is supported from research-learn by providing the
ModelSearchCV
class and the report_model_search_results()
function.
Design and analysis of experiments¶
Setting up and analyzing the results of machine learning experiments
is a time consuming procedure that requires multiple steps like
data gathering and preparation, selection of estimators and their
hyperparameters, analysis of the results and finally application of
multiple statistical tests. The application of these steps is simplified
by research-learn through the base BaseExperiment
as well as
appropriate functions to analyze the experimental results.