reward uncertainty: Nonlinear Function