Keywords: learning probability threshold; pattern recognition; classification costs; medical diagnosis; learning optimal decision scheme; two-class two-action case; minimum risk classification; recursive estimation; convergence in probability; optimum threshold