Pearson Correlations on Networks: Addendum
At the link above you will find a zipfile containing the code and data to replicate the claims in “Pearson Correlations on Networks: Constraints and Solutions,” a research note motivating the usage of effective resistance or node embeddings as the distance matrix to calculate proper network correlations — moving away from the suggestion in the main paper to use shortest path distance.
Running the two Jupyter notebooks should be sufficient. One requires a Julia 1.8.0 kernel, the other a Python 3 one. They will produce the same results (hopefully!).
You will need the following Julia packages to run the Julia code: Laplacians, Graphs, LinearAlgebra, Statistics, Word2Vec, DelimitedFiles, Distances.
To run the Python code, these packages will be necessary instead: numpy, scipy, gensim, sklearn, networkx, pandas, pyemd.
