On the \(L_{2}\)-discrepancy (Q2381825): Difference between revisions

In this article, the following problem is considered. Let \(\mathcal{S}\) be a family of subsets of a finite set \(V\) and let \(\chi:V\rightarrow \{-1,1\}\) be a coloring of \(V\). For \(S_j\in\mathcal{S}\), let \(\chi(S_j):=\sum_{v\in S_j}\chi (v)\) and define the \(L_2\)-discrepancy of \(\mathcal{S}\) with respect to \(\chi\) as \[ \text{disc}_2(\mathcal{S},\chi):=\biggl(\frac{1}{| \mathcal{S} |}\sum_j |\chi (S_j)|^2\biggr)^{1/2}. \] Furthermore, the \(L_2\)-discrepancy of \(\mathcal{S}\) is defined by \[ \text{disc}_2 (\mathcal{S}):=\min_\chi \text{disc}_2 (\mathcal{S},\chi), \] where the minimum is extended over all colorings \(\chi\). The authors of this note discuss a relation of the \(L_2\)-discrepancy problem to the so called MAX-CUT problem and show two main results. Firstly, it is shown that the \(L_2\)-discrepancy problem is NP-complete. Moreover, the authors show that there exists a coloring \(\chi\) such that \[ \biggl(\frac{1}{| \mathcal{S} |}\sum_j |\chi (S_j)|^2\biggr)^{1/2}\leq\sqrt{| V|}. \]