Export Reviews, Discussions, Author Feedback and Meta-Reviews

I do not see how this can be achieved as long as the k-neighborhoods and the pooling scale, given by the covariance matrix Sigma, are shared across all feature maps.