Asymptotic behavior of the loss function in the multiplicative scaling of the wavelet coefficients of a signal function

Abstract

The construction of an estimator for a signal function from noisy data is considered. The technique depends on minimizing the probability of exceeding the maximum error in calculating the wavelet coefficients for a given critical value. In a model with additive Gaussian noise, the order of the loss function is estimated in a class of Lipschitz regular functions.