``Curse of dimensionality'' for complexity of approximation for classes of functions satisfying Lipschitz condition (Q1377427)

A theorem is proved which particularly implies that the least number of operations of summing, extraction and multiplication is necessary to calculate at every point of the unit \(n\)-dimensional cube the values of functions uniformly approximated on this cube with error not bigger than \(\varepsilon\), the given function satisfying a Lipschitz condition in the \(i\)-th variable with constant \(M_i\) and bounded by modulo constant \(N\), does not asymptotically exceed \(H/\log_2h\), where \(H = 3(n/2\varepsilon)^n \prod\limits_{i=1}^n M_i\) as \(\varepsilon \to 0\), bounded (not vanishing) constants \(M_i\) and \(N\) and for some \(n\), both bounded and tending to infinity. The entropy lower estimation of the considered complexity of \(\varepsilon\)-approximation of the whole class of functions satisfying the above conditions is asymptotically \((2e)^n\) smaller than the cited upper estimate.