Adaptive wavelet density estimation under independence hypothesis (Q2242440)

In the paper, an estimator on isotropic Besov spaces over \(\mathbb R^d\) is derived. It is a data-driven wavelet estimator, based on an orthonormal scaling function. Since wavelet analysis provides local analysis and fast algorithm, the convergence rates are very good. In the case of Nikol'skii class \(\mathcal N_r(s,\mathbb R^d)\), which is equal to Besov space \(B_{r,\infty}^s(\mathbb R^d)\), the authors obtain the already known upper bounds. In the case of density functions satisfying the independence hypothesis, convergence rates are better than that derived previously. The paper is well-organized and well-written, it contains an extensive literature survey. All the necessary notions are introduced. The proofs are technical. Linguistic mistakes slightly disturb reading.