Markov Blanket Ranking using Kernel-based Conditional Dependence Measures
This is an incremental improvement for scientists needing more accurate causal feature selection from observational data.
The paper tackled the limitation of forward selection in Markov blanket discovery by proposing a backward elimination method using kernel-based conditional dependence measures, which improved performance on synthetic and real datasets.
Developing feature selection algorithms that move beyond a pure correlational to a more causal analysis of observational data is an important problem in the sciences. Several algorithms attempt to do so by discovering the Markov blanket of a target, but they all contain a forward selection step which variables must pass in order to be included in the conditioning set. As a result, these algorithms may not consider all possible conditional multivariate combinations. We improve on this limitation by proposing a backward elimination method that uses a kernel-based conditional dependence measure to identify the Markov blanket in a fully multivariate fashion. The algorithm is easy to implement and compares favorably to other methods on synthetic and real datasets.