Deconvolution of k-monotone densities
vendredi 3 octobre 2014, 10h30 - 11h30
In this talk, we discuss the problem of statistical inference for a k-monotone density based on a sample from it corrupted by noise.
We derive the minimax convergence rates, construct a kernel estimate which achieves this rates and discuss the optimality of the rates.
The construction of the corresponding lower bounds is closely connected to the so called uncertainty principle in Fourier analysis.