Learning Markov Network Structure using Brownian Distance Covariance
This addresses the challenge of structure learning in Markov networks for researchers in statistics and machine learning, though it appears incremental as it adapts an existing covariance measure to this domain.
The paper tackles the problem of learning undirected graph structures from data by proposing a non-parametric method using Brownian distance covariance to estimate conditional independences, applicable in high-dimensional settings where parameters exceed sample size.
In this paper, we present a simple non-parametric method for learning the structure of undirected graphs from data that drawn from an underlying unknown distribution. We propose to use Brownian distance covariance to estimate the conditional independences between the random variables and encodes pairwise Markov graph. This framework can be applied in high-dimensional setting, where the number of parameters much be larger than the sample size.