收录:
摘要:
Donoho et al. in 1996 have made almost perfect achievements in wavelet estimation for a density function f in Besov spaces B-r,q(s)(R). Motivated by their work, we define new linear and nonlinear wavelet estimators f(n,m)(lin), f(n,m)(non) for density derivatives f((m)). It turns out that the linear estimation E(parallel to f(n,m)(lin) - f((m))parallel to(p)) for f((m)) is an element of B-r,q(s)(R) attains the optimal when r >= p, and the nonlinear one E(parallel to f(n,m)(non) - f((m))parallel to(p)) does the same if r <= p/2(s+m)+1. In addition, our method is applied to Sobolev spaces with non-negative integer exponents as well.
关键词:
通讯作者信息: