Validating reconstructors against reference implementations

[sdmccabe]

Same as #245 but for the reconstructors. My suspicion is that most of these don't have easily comparable reference implementations, so those can be checked off fairly quickly. The implementers of each method probably did some checks for reference implementations so we should reach out to them.

• [ ] Convergent Cross Mapping
• [ ] Correlation Matrix
• [ ] Correlation Spanning Tree
• [ ] Free Energy Minimization
• [ ] Graphical Lasso
• [ ] Marchenko Pastur
• [ ] Maximum Likelihood Estimation
• [ ] Mean Field
• [ ] Mutual Information Matrix
• [ ] Naive Transfer Entropy
• [x] Optimal Causation Entropy
• [ ] OU Inference
• [ ] Partial Correlation Influence
• [ ] Partial Correlation Matrix
• [x] Random
• [ ] Thouless Anderson Palmer
• [ ] Time Granger Causality

[leotrs]

It'd be nice (if at all possible) to have a single script/notebook to run once we have identified reference implementations.

[sdmccabe]

I was thinking about something similar; the challenge of course is that not all the reference implementations are in Python. I guess for R-based implementations we can still use Jupyter, but what about Matlab or C++? Another possibility would be another repository containing the validation scripts.

[leotrs]

Uuuuuuh let's not do another repo...