about equation 1 in paper

[zackzhao1]

why is div m^2？maybe C(2，m)，Σ/(m(m-1)/2) as the code in this git:

eucl_dist = euclidean_distances(norm, norm)[np.triu_indices(T, k=1)]   
return embedding, 2*(1/(1+np.exp(np.mean(eucl_dist))))

[pterhoer]

Hi zackzhao1,

Thank you for this comment. You are right.
You can fix that in mutliple ways. One is by changing the sum to go over all pairs (sum_{i,j}).
Practically, this minor mistake will not change the performance of the quality predictions since it is a constant factor that is applied to every quality prediction :)

Best regards
Philipp