人工智能培训

搜索

深度学习论文:高效的基于搜索的加权模型集成(Efficient Search-Based Weighted Model Integration)

[复制链接]
zycwolf 发表于 2019-3-15 13:18:54 | 显示全部楼层 |阅读模式
zycwolf 2019-3-15 13:18:54 148 0 显示全部楼层
深度学习论文:高效的基于搜索的加权模型集成(Efficient Search-Based Weighted Model Integration)加权模型集成(WMI)将加权模型计数(WMC)扩展到混合离散连续域上的函数集成。它在解决图形模型和概率编程中的推理问题方面具有显着的前景。然而,WMI的最先进工具是有限的性能间隔,忽略了对提高效率至关重要的独立结构。为了解决这个局限性,我们提出了一种有效的模型积分算法,用于具有树原始图的理论。我们通过使用搜索来执行集成来利用稀疏图结构。我们的算法大大提高了这些问题的计算效率,并利用了变量之间的特定于文本的独立性。实验结果表明,与现有的WMI求解器相比,树形依赖性问题的实际加速比较。
Weighted model integration (WMI) extends Weighted model counting (WMC) to theintegration of functions over mixed discrete-continuous domains.It has showntremendous promise for solving inference problems in graphical models andprobabilistic programming.Yet, state-of-the-art tools for WMI are limited interms of performance and ignore the independence structure that is crucial toimproving efficiency.To address this limitation, we propose an efficient modelintegration algorithm for theories with tree primal graphs.We exploit thesparse graph structure by using search to performing integration.Our algorithmgreatly improves the computational efficiency on such problems and exploitscontext-specific independence between variables.Experimental results showdramatic speedups compared to existing WMI solvers on problems with tree-shapeddependencies.深度学习论文:高效的基于搜索的加权模型集成(Efficient Search-Based Weighted Model Integration) d7P9dE9hwStRz9wx.jpg
URL地址:https://arxiv.org/abs/1903.05334     ----pdf下载地址:https://arxiv.org/pdf/1903.05334    ----深度学习论文:高效的基于搜索的加权模型集成(Efficient Search-Based Weighted Model Integration)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

zycwolf当前离线
新手上路

查看:148 | 回复:0

快速回复 返回顶部 返回列表