Diffusion operator construction

[qubit0]

I am trying to understand how the diffusion operator is constructed. I can follow up to cell-cell distance matrix i.e.

After that affinity matrix caculated using a Gaussian kernel. Here, we have to choose sigma.

1. How is sigma chosen?
2. Where in the python source code, can I find the adapative tuning of sigma?
3. What is magic_operator.graph.kernel ? How is it constructed? If we set non-zero values to zero here than does it become an adjacency matrix for the graph?
4. magic_operator.graph.diff_op is a row normalized Markov matrix M(i,j), right ?