Lifted Probabilistic Inference: An MCMC Perspective

Lifted Probabilistic Inference: An MCMC Perspective
复制标题

提升概率推理:MCMC 视角

DOI:
--
复制
发表时间:
2012
期刊:
StarAI@UAI
影响因子:
--
通讯作者:
Mathias Niepert
Mathias Niepert
中科院分区:
--
文献类型:
--
作者:
Mathias Niepert

文献摘要

被引文献

相似文献

普遍的共识似乎是解除了 推理与利用模型有关 对称性和分组无法区分 推理时的对象。从一阶开始 概率形式主义本质上是临时的 板语言提供更紧凑 相应地面的表示 模型中,提升推理往往在这些模型中效果特别好。我们证明了 不可区分性的概念本身就显现出来 在几个不同的水平上{常数的水平,基本原子(变量)的水平, 公式(特征)的级别和级别 任务(可能的世界)。我们讨论 MCMC 文献中关于前的现有工作 利用变量水平上的对称性 作业并将其与新结果联系起来 解除了MCMC。
The general consensus seems to be that lifted inference is concerned with exploiting model symmetries and grouping indistinguishable objects at inference time. Since first-order probabilistic formalisms are essentially tem- plate languages providing a more compact representation of a corresponding ground model, lifted inference tends to work especially well in these models. We show that the notion of indistinguishability manifests itself on several dferent levels {the level of constants, the level of ground atoms (variables), the level of formulas (features), and the level of assignments (possible worlds). We discuss existing work in the MCMC literature on ex- ploiting symmetries on the level of variable assignments and relate it to novel results in lifted MCMC.