Codes for performing the Relevant Information Sampling approach
In this study, we propose a novel strategy to address the class imbalance issue using a Relevant Information-based sampling approach. In this sense, our RIS methodology employs a principle of relevant information cost function which formulates the sub-sampling issue as a trade-off between the entropy
Parameters: X : Majority Class.
N_size : Samples to select from the majority Class. Generally equal to the size of the minority class.
lambda : The weighting parameter which combines the regularization and similarity terms in appropriate proportion to capture different aspects of the data structure.
sigma : The resolution parameter which controls the “scale” of the analysis
Copyright (C) 2018 Automatica Research Group
Keider Hoyos Osorio -Universidad Tecnológica de Pereira