arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1803.01422stat.MLcs.AIcs.LGstat.ME
收藏

具有NO TEARS的DAG:结构学习的连续优化

DAGs with NO TEARS: Continuous Optimization for Structure Learning

Carnegie Mellon University(卡内基梅隆大学)

Xun Zheng, Bryon Aragam, Pradeep Ravikumar, Eric P. Xing

摘要:

估计有向无环图(DAG,也称为贝叶斯网络)的结构是一个具有挑战性的问题,因为DAG的搜索空间是组合的,并且与节点的数量成超指数关系。现有的方法依赖于各种本地算法来执行非循环约束。在本文中,我们引入了一个根本不同的策略:我们制定的结构学习问题作为一个纯粹的\emph{continuous}优化问题,真正的矩阵,完全避免了这种组合约束。这是通过一种新的非周期性的表征,不仅是顺利的,而且准确。由此产生的问题可以有效地解决标准的数值算法,这也使得实施毫不费力。所提出的方法优于现有的,而不施加任何结构上的假设,如有界树宽或度的图。实现该算法的代码是开源的,可在https://github.com/xunzheng/notears上公开获得。

英文摘要:

Estimating the structure of directed acyclic graphs (DAGs, also known as Bayesian networks) is a challenging problem since the search space of DAGs is combinatorial and scales superexponentially with the number of nodes. Existing approaches rely on various local heuristics for enforcing the acyclicity constraint. In this paper, we introduce a fundamentally different strategy: We formulate the structure learning problem as a purely \emph{continuous} optimization problem over real matrices that avoids this combinatorial constraint entirely. This is achieved by a novel characterization of acyclicity that is not only smooth but also exact. The resulting problem can be efficiently solved by standard numerical algorithms, which also makes implementation effortless. The proposed method outperforms existing ones, without imposing any structural assumptions on the graph such as bounded treewidth or in-degree. Code implementing the proposed algorithm is open-source and publicly available at https://github.com/xunzheng/notears.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑