Export Reviews, Discussions, Author Feedback and Meta-Reviews

In the synthetic experiments, it was chosen such that the F-score is maximized. Forcing TiWnet to produce a more sparse topology might appear logical at first, but this will equally remove false _and_ true positive edges. In TiWnet's model, feature independence is hard-coded by design, hence it is forced to explain all observations by structure only. It cannot distinguish these two sources of input.