Minimizing the \(L_{2}\) and \(L_{\infty}\) star discrepancies of a single point in the unit hypercube (Q2432729)

The authors consider the so-called \(L_{\infty}\) star discrepancy and \(L_{2}\) star discrepancy of an arbitrary point set, composed of \(N\) points in \(s\)-dimensional hypercube \([0,1]^{s}\) as quantitative measures for the quality of the distribution of this point set. For a point set that contains only a single point \({\mathbf x}\) in \([0,1]^{s}\) the authors show where to place the point \({\mathbf x}\) in order to minimize these measures and how to calculate these minimums. Theorem 2 states that for a point set \(Q\) that contains only one \(s\)-dimensional point, the minimum \(L_{\infty}\) star discrepancy is reached for a point on the main diagonal of \([0,1]^{s},\) so the point \({\mathbf x}\) is of the form \({\mathbf x} = (x, \ldots ,x),\) where \(x\) is the only root in \([0,1]\) of the polynomial \(x^{s} + x - 1.\) The minimum value of the \(L_{\infty}\) star discrepancy \( D_{1}^{*}(\{ {\mathbf x} \}) = \max (x, 1-x^{s}) \) is obtained. A table 1 lists minima of \(D_{1}^{*}(\{ {\mathbf x} \})\) for dimensions \(s = 2^{k}, k= 1, \dots ,6.\) Theorem 4 states that for a point set \(Q\) that contains only one \(s\)-dimensional point, the minimum \(L_{2}\) star discrepancy is reached for a point on the main diagonal of \([0,1]^{s},\) so the point \({\mathbf x}\) is of the form \({\mathbf x} = (x, \dots ,x),\) where \(x\) is the only root in \([0,1]\) of the polynomial \(x (1+x)^{s-1} - 2^{s-2}.\) It is shown that the minimum \(L_{2}\) star discrepancy \(T_{1}^{*}(\{ {\mathbf x} \})\) converges to \(3^{-s/2}\) when the dimension \(s \to \infty.\) A table 1 also lists minima of \(T_{1}^{*}(\{ {\mathbf x} \})\) and the values of \(\sqrt{3^{-s}}\) for dimensions \(s = 2^{k}, k= 1, \ldots ,6.\)