The use of nonparametric approaches and semiparametric approaches for modeling fatigue and performance are analyzed. Nonparametric approaches in the form of stand-alone artificial neural networks and semiparametric (hybrid) approaches that combine neural networks with prior process knowledge are explored and compared with existing parametric approaches based on the two-process model of sleep regulation. Within the context of a military application, we explore two notional semiparametric approaches for real-time prediction of cognitive performance on the basis of individualized on-line measurements of physiologic variables. Initial analysis indicates that these alternative modeling approaches may address key technological gaps and advance fatigue and performance modeling. Most notably, these approaches seem amenable to predicting individual performance and quantitatively assessing the reliability of model predictions through estimation of statistical error bounds, which have eluded researchers for the last two decades.