Towards Compact Interpretable Models : Shrinking of Learned Probabilistic Sentential Decision Diagrams
Towards Compact Interpretable Models : Shrinking of Learned Probabilistic Sentential Decision Diagrams
复制标题
走向紧凑的可解释模型：学习概率句子决策图的收缩
DOI：
--
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Rain.
中科院分区：
文献类型：
--
作者：
Sun;Rbow;Rain.
最近将概率句子决策图（PSDD）作为离散概率分布的可解释和可解释的表示。 PSDD是可以进行的，因为它们有效地支持了广泛的查询。它们是可以解释的，因为PSDD中的每个参数都表示有条件的概率，如贝叶斯网络中。本文总结了正在进行的研究，旨在回答两个对使用PSDD作为可解释的AI模型很重要的问题。首先，作为一种可解释的模型，PSDD可以与更通用的机器学习模型竞争密度估计吗？我们积极回答这个问题，报告了标准基准测试的最新结果。其次，我们是否可以有效地减少学习的PSDD中参数的数量，以简化其解释，而不会损害学习模型的质量？对于此任务，我们提出了一种算法，该算法合并了KL-Divergence中相似的PSDD子结构，我们显示的可以在PSDD上有效地完成。
Probabilistic sentential decision diagrams (PSDDs) were recently introduced as a tractable and interpretable representation of discrete probability distributions. PSDDs are tractable because they support a wide range of queries efficiently. They are interpretable because each parameter in the PSDD represents a conditional probability, as in Bayesian networks. This paper summarizes ongoing research that aims to answer two questions that are important to employ PSDDs as an explainable AI model. First, as a tractable and interpretable model, can PSDDs compete with more general machine learning models for density estimation? We answer this question positively, reporting state-of-the-art results on standard benchmarks. Second, can we effectively reduce the number of parameters in a learned PSDD to simplify its interpretation, without harming the quality of the learned model? For this task, we present an algorithm that merges PSDD substructures that are similar in KL-divergence, which we show can be done efficiently on PSDDs.
DOI：
--
发表时间：
2017
期刊：
Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence (UAI
影响因子：
--
作者：
Liang, Yitao;Bekker, Jessa;Van den Broeck, Guy
通讯作者：
Van den Broeck, Guy