MAP Inference with MRF by Graduated Non-Convexity and Concavity Procedure

	MAP Inference with MRF by Graduated Non-Convexity and Concavity Procedure
	Zhi-Yong Liu; Hong Qiao; Jian-Hua Su
	2014
会议名称	Neural Information Processing 21st International Conference, ICONIP 2014
会议日期	3-6 Nov. 2014
会议地点	Kuching, Malaysia
关键词	NONE
通讯作者	Zhi-Yong Liu
英文摘要	In this paper we generalize the recently proposed graduated non-convexity and concavity procedure(GNCCP) to approximately solve the maximum a posteriori (MAP) inference problem with the Markov random field (MRF). Unlike the commonly used graph cuts or loopy brief propagation, the GNCCP based MAP algorithm is widely applicable to any types of graphical models with any types of potentials, and is very easy to use in practice. Our preliminary experimental comparisons witness its state-of-the-art performance.
会议录	Neural Information Processing. 21st International Conference, ICONIP 2014. Proceedings: LNCS 8835
内容类型	会议论文
源URL	[http://ir.ia.ac.cn/handle/173211/12866]
专题	自动化研究所_复杂系统管理与控制国家重点实验室_机器人应用与理论组
推荐引用方式 GB/T 7714	Zhi-Yong Liu,Hong Qiao,Jian-Hua Su. MAP Inference with MRF by Graduated Non-Convexity and Concavity Procedure[C]. 见:Neural Information Processing 21st International Conference, ICONIP 2014. Kuching, Malaysia. 3-6 Nov. 2014.