## Short-term
- Record run times of each function call in demo GMMN_MTS_paper.R
- Run this demo again on Graham

## Long-term
- Fix: Paper citations (2019 on arXiv or the real publication if published)
- KL(): If an analytical (differentiable) form exits for K-L divergence
        between copulas, we could propose an interesting extension.
        (Simpler) extensions to consider first: encoder maps to non-diagonal
        covariance matrix, Kendall matrix (which we then transform to covariance
        matrix for use in KL()).
- Need:
  1) Functions to extract parameters of trained GNNs and convert them to a NN in pure R
     code (so that one can call them in parallel)
  2) Perhaps even implement fitting/training in R (even if inefficient)
