We investigate a refined recursive coloring approach to construct balanced colorings for hypergraphs. A coloring is called balanced if each hyperedge has (roughly) the same number of vertices in each color. We provide a recursive randomized algorithm that colors an arbitrary hypergraph (n vertices, m edges) with c colors with discrepancy at most O(the square root of (n/c log m)). The algorithm has expected running time O(nm log c). This result improves the bound of O (the square root of (n log (cm)) achieved with probability at least 1/2 by a random coloring that independently chooses a random color for each vertex (fair dice coloring). Our approach also lowers the current best upper bound for the c-color discrepancy in the case n = m to O (the square root of (n/c log c)) and extends the algorithm of Matousek, Welzl and Wernisch for hypergraphs having bounded dual shatter function to arbitrary numbers of colors.
展开▼
机译:我们调查一种精致的递归着色方法来构建超图的平衡着色。如果每个HINFEGPE(大致)每种颜色中相同数量的顶点,则称为彩色。我们提供递归随机算法,其颜色与C颜色的任意超图(n顶点,m边缘)颜色为大多数差异((n / c log m)的平方根)。该算法预计运行时间O(nm log c)。该结果改善了通过随机着色的概率至少1/2实现的O(n log(cm)的平方根)的界限,其独立选择每个顶点的随机颜色(公平骰子着色)。我们的方法也降低了C颜色差异的当前最佳上限为n = m到o((n / c log c)的平方根)并扩展了Matousek,Welzl和Wernisch的算法,用于具有有界双碎函数的超图任意数量的颜色。
展开▼