Asymptotically linear estimation in a generalized linear model

Suppose that Y = f(X(tau)beta)+epsilon Here f is a smooth but unknown function, beta is a k x 1 parameter vector to be estimated and epsilon(i) is a random error with mean 0 and variance sigma(2). The asymptotically linear estimator of beta is constructed based on the model Y-i = f(X(i)(tau)beta)+epsilon(i), i = 1 ... n, when the density functions of epsilon and X are unknown.