Dear list, what exactly does the 'recs' (and rec_types) parameter in 'BlockState' and 'LayeredBlockState' do? I couldn't find an example in the documentation. But for instance if I have a network with integer edge weights (stored in the PropertyMap 'weight'), I could specify recs=[weight], rec_types=["discrete-poisson"]. And probably that would assume a Poisson distribution for the weights; am I right? If I am, then would this be the same as specifying eweight=weight and deg_corr=False? Many thanks in advance, Peter
On 08.03.2017 07:35, Peter Straka wrote:
Dear list,
what exactly does the 'recs' (and rec_types) parameter in 'BlockState' and 'LayeredBlockState' do?
I couldn't find an example in the documentation. But for instance if I have a network with integer edge weights (stored in the PropertyMap 'weight'), I could specify recs=[weight], rec_types=["discrete-poisson"]. And probably that would assume a Poisson distribution for the weights; am I right?
You are right, but the reason why this is not yet documented in detail is because it is still work in progress.
If I am, then would this be the same as specifying eweight=weight and deg_corr=False?
Not quite, because the prior probabilities for the parameters are specified in a different way. It will all be explained in the documentation once it is ready. Best, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>
Great, thank you Tiago, I will just use the eweight parameter then. Peter On Wed, 8 Mar 2017 at 22:42 Tiago de Paula Peixoto <tiago@skewed.de> wrote:
On 08.03.2017 07:35, Peter Straka wrote:
Dear list,
what exactly does the 'recs' (and rec_types) parameter in 'BlockState' and 'LayeredBlockState' do?
I couldn't find an example in the documentation. But for instance if I have a network with integer edge weights (stored in the PropertyMap 'weight'), I could specify recs=[weight], rec_types=["discrete-poisson"]. And probably that would assume a Poisson distribution for the weights; am I right?
You are right, but the reason why this is not yet documented in detail is because it is still work in progress.
If I am, then would this be the same as specifying eweight=weight and deg_corr=False?
Not quite, because the prior probabilities for the parameters are specified in a different way. It will all be explained in the documentation once it is ready.
Best, Tiago
-- Tiago de Paula Peixoto <tiago@skewed.de>
_______________________________________________ graph-tool mailing list graph-tool@skewed.de https://lists.skewed.de/mailman/listinfo/graph-tool
participants (2)
-
Peter Straka -
Tiago de Paula Peixoto