Jean-Marc Brossier

Analytical study of the optimal combination of binary classifiers based on classifiers-induced partitioning of the training set

Jean-Marc Brossier, Olivier Lafitte

This paper studies an optimal linear combination of binary classifiers based on a logical structuration of the dataset via truth tables. The given classifiers partition data into equivalence classes, allowing for a rigorous analysis of the convexified empirical risk through a mul…Read more