OPTIMAL SIGNAL ESTIMATION USING CROSS-VALIDATION

This letter develops an optimal, nonlinear estimator of a deterministi c signal in noise, The methods of penalized least-squares and cross-va lidation (CV) balance the bias-variance tradeoff and lead to a closed form expression for the estimator, The estimator is simultaneously opt imal in a ''small-sample,'' predictive sum of squares sense and asympt otically optimal in the mean square sense.