Authors: Xiaoyan Zhang ^1,2; Zan-Bo Zhang ^2,3; Hajo Broersma ²; Xuelian Wen ⁴

• 1 School of Mathematical Science & Institute of Mathematics [Nanjing]
• 2 Faculty of Electrical Engineering, Mathematics and Computer Science [Twente]
• 3 Department of Computer Engineering, Guangdong Industry Technical College
• 4 School of Economics and Management, South China Normal University

Network models allow one to deal with massive data sets using some standard concepts from graph theory. Understanding and investigating the structural properties of a certain data set is a crucial task in many practical applications of network optimization. Recently, labeled network optimization over colored graphs has been extensively studied. Given a (not necessarily properly) edge-colored graph $G=(V,E)$, a subgraph $H$ is said to be <i>monochromatic</i> if all its edges have the same color, and called <i>multicolored</i> if all its edges have distinct colors. The monochromatic clique and multicolored cycle partition problems have important applications in the problems of network optimization arising in information science and operations research. We investigate the computational complexity of the problems of determining the minimum number of monochromatic cliques or multicolored cycles that, respectively, partition $V(G)$. We show that the minimum monochromatic clique partition problem is APX-hard on monochromatic-diamond-free graphs, and APX-complete on monochromatic-diamond-free graphs in which the size of a maximum monochromatic clique is bounded by a constant. We also show that the minimum multicolored cycle partition problem is NP-complete, even if the input graph $G$ is triangle-free. Moreover, for the weighted version of the minimum monochromatic clique partition problem on monochromatic-diamond-free graphs, we derive an approximation algorithm with (tight) approximation guarantee ln $|V(G)|+1$.