Approximation of anisotropic classes by standard information (Q2483207)

Let \(\Delta_\sigma(K,X)\) denote the minimum information diameter of the set in the space \(X(\mathbb{R}^d)\), \(E_\sigma(K,X)\) do the minimum intrinsic error of the optimal recovery of \(K\) in \(X\), and \(E^L_\sigma(K,X)\) be the minimum linear intrinsic error. Then it well known that \[ {1\over 2}\Delta_\sigma(K,X)\leq E_\sigma(K,X)\leq E^L_\sigma(K, X). \] In particular, take \(K\) as the unit ball of Besov space \(B^r_{p,\theta}(\mathbb{R}^d)\), and \(X\) as the Sobolev space \(L^s_q(\mathbb{R}^d)\). Then, under some technical assumptions of \(p\), \(q\), \(r\), \(\theta\), this paper gives a lower bound \(C_1\sigma^{-a(\omega-\eta)}\) of \(\Delta_\sigma(K,X)\) and an upper bound \(C_2\sigma^{-a(\omega-\eta)}\) of \(E^L_\sigma(K,X)\). Here, \(a\), \(\omega\), \(\eta\) are constants which are determined by \(p\), \(q\), \(r\).