Similarity measures for networks with same nodes

Question

[nyanovsky]

I'm simulating a GRN and using an experimental one as validation. What I want to do is compare the simulated one with the experimental one as I tune some parameter of my algorithm in the recreation of the network, and see what values of that parameter achieve a network that is closest to the experimental one. What are some good similarity measures I can use to make this comparison, taking into account that both networks have the same nodes?

Answer 1

[LChart]

You could use the Jaccard similarity of the edge-sets; or for a weighted network the cosine similarity of the edge weights; you could use the Frobenius norm as your metric and compute $||adjacency(A)-adjency(B)||_F$.