In this paper, we propose a high-speed greedy sequential algorithm for the vertex coloring problem (VCP), based on the Wave Function Collapse algorithm, called Wave Function Collapse Coloring (WFC-C). An iteration of this algorithm goes through three distinct phases: vertex selection, color restriction through wave function collapsing, and domain propagation. In effect, WFC-C propagates color choices or "domain" restrictions beyond immediate neighbourhoods. This heuristic, combined with a series of other greedy optimizations, allows for a fast algorithm that prevents certain color conflicts. Through extensive experiments, we show that WFC-C remains competitive (and occasionally better) in terms of optimal coloring, and dramatically outperforms existing high-speed VCP, with on average speed differences ranging from 2000 times to 16000 times, on the most difficult instances of the DIMACS benchmark.
翻译:在本文中,我们基于波形函数折叠算法(Wave 函数折叠算法,称为Wave 函数折叠颜色(WFC-C)),为顶层颜色问题提出了一个高速贪婪序列算法(VCP ) 。 这一算法的循环经历了三个不同的阶段:顶层选择、波形函数折叠时的颜色限制以及域传播。实际上,WFC-C将颜色选择或“主”限制推广到近邻之外。这一超常加上一系列其他贪婪优化,使得快速算法能够防止某些颜色冲突。通过广泛的实验,我们证明WFC-C在最佳颜色化方面仍然具有竞争力(有时会更好 ), 并且大大超过现有的高速VCP, 平均速度差异从2000倍到16000倍不等, 是在DIMACS 基准的最困难的案例中。