an implementation of the NO-TEARS Bayesian network optimization algorithm in Tensorflow 2.

A Python package for plug-and-play DAG and Bayesian network optimization using the NO-TEARS algorithm.

NO-TEARS is a unique use of the matrix exponential that creates a differentiable scoring function for the acyclicity of a graph, providing the first smooth loss for optimization-based learning of DAGs and Bayesian networks. This is the bedrock for my work on meta-learning approaches for personalized Bayesian network inference.