The greedy L2L_2 discrepancy conjecture for the sequence generated by Algorithm 1

From papers

Let S\mathcal{S}^* be the sequence generated by Algorithm 1, and let V~\widetilde{\mathcal{V}} be the symmetrized van der Corput sequence in base 22. Write L2,N(S)L_{2,N}(\mathcal{S}) for the L2L_2 discrepancy of the first NN elements of a sequence S\mathcal{S}.

Greedy L2L_2 discrepancy conjecture. The sequence S\mathcal{S}^* should satisfy

lim supNL2,N(S)logN<lim supNL2,N(V~)logN0.319553.\limsup_{N\to\infty}\frac{L_{2,N}(\mathcal{S}^*)}{\sqrt{\log N}}<\limsup_{N\to\infty}\frac{L_{2,N}(\widetilde{\mathcal{V}})}{\sqrt{\log N}}\leq 0.319553\ldots.

Numerical experiments indicate that S\mathcal{S}^* generally has smaller L2L_2 discrepancy than the symmetrized van der Corput sequence, while the displayed upper bound for the latter is attributed to Faure. An explicit description of S\mathcal{S}^* and sharper discrepancy bounds remain desirable.

Progress summary

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Sources & referencesView supporting material

Primary source

Ralph Kritzinger, “Uniformly distributed sequences generated by a greedy minimization of the L_2 discrepancy”, arXiv:2109.06298 (2022).

Solutions 0

No solutions have been posted yet.