FA(IR)²MA-GLVQ – A hidden-feature-bias mitigation approach for fairness in classification learning based on generalized matrix learning vector quantization Journalartikel