This is "harmonic" centrality metric realization for networkx library.
It uses HyperLogLog algorithm and much more faster than standart algorithm from networkx.
Harmonic centrality is calculated as 
Harmonic centrality is often better than classic centrality algorithms(PageRank, Katz, Closeness, Betweenness). See this: https://events.yandex.ru/lib/talks/1287/ video for details
installing:
pip install git+https://github.com/asash/harmonic_centrality.git
example usage:
from networkx import Graph
from harmonic_centrality import harmonic_centrality
G = Graph()
G.add_edge(1,2)
G.add_edge(1,3)
G.add_edge(1,4)
G.add_edge(2,3)
harmonic = harmonic_centrality(G)
for node in harmonic:
print node, harmonic[node]output:
1 4.33868071575
2 3.61485913805
3 3.61485913805
4 2.89174614742