Hi Tiago,
From the documentation of deg_sampler: "This function is called once per vertex, but may be called more times, if the degree sequence cannot be used to build a graph."
Now suppose my deg_sampler sometimes returns values greater than N-1, and if I don't want to generate a graph with multi-edges and self-loops, such values will be discarded. But suppose for first few vertices, drawn values were less than N (and hence are accepted) and the next value is greater than N-1. Now will all the values generated so far discarded or only the last value? I feel that discarding only the last value will create a bias if I want to sample degrees from a particular probability distribution. Could you please clarify this? Thank you SS ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Thursday, July 23, 2020 2:09 PM, Snehal Shekatkar <snehalshekatkar@protonmail.com> wrote:
Thanks so much Tiago!
Sent with ProtonMail Secure Email.
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Thursday, July 23, 2020 1:35 PM, Tiago de Paula Peixoto tiago@skewed.de wrote:
Am 23.07.20 um 08:31 schrieb Snehal Shekatkar:
Sorry for bothering again. A small query: If the graph size is say 10^4, and the degrees are drawn from the discrete-power law or some other right-skewed distribution for which the second moment diverges, would n_iter = 1000 be enough for the Markov chain to saturate? Is there a rule of thumb for choosing n_iter when the scaling index of the power-law and the graph size are given?
Unfortunately, there is no known rule of thumb to know how fast the chain mixes. My inclination is to say that 1000 sweeps is often enough, but I would encourage you to experiment and draw your own conclusions.
Tiago de Paula Peixoto tiago@skewed.de graph-tool mailing list graph-tool@skewed.de https://lists.skewed.de/mailman/listinfo/graph-tool