无需有向非循环图的因果学习
因果学习方法通常根据其发现真实底层有向非循环图(DAG)结构的能力来评估。
Causal learning methods are often evaluatedin terms of their ability to discover a true underlying directed acyclic graph(DAG) structure.
然而,一般来说,真正的结构是未知的,可能并不是DAG结构
However, in general the true structure isunknown and may not be a DAG structure.
因此,我们考虑在未知测试数据上根据干预效果的预测评估因果学习方法。
We therefore consider evaluating causallearning methods in terms of predicting the effects of interventions on unseentest data.
基于此任务,我们的研究表明,各种各样的因果关系建模方法都可以泛化为基于DAG的方法。
Given this task, we show that there exist avariety of approaches to modeling causality, generalizing DAG-based methods.
我们在合成生物学数据上的实验表明,一些非DAG模型在因果预测任务上比基于DAG的方法表现得更好。
Our experiments on synthetic and biologicaldata indicate that some non-DAG models perform as well or better than DAG-basedmethods at causal prediction tasks.
使用有向非循环图(DAG)设计因果模型是很常见的。
It is common to make causal models usingdirected acyclic graphs (DAGs).
然而,这种方法的一个问题是,很难评估图形结构是否正确。
However, one problem with this approach isthat it is very hard to assess whether the graph structure is correct or not.
……
下载英文原文及源代码地址:
http://page2.dfpan.com/fs/bl4ccje272012209160/
更多精彩文章请关注微信号: