Complete the sentence for GLASSO: … The large the lambda the …
fewer edeges the final network has
Whats the principal idea behind stability selection for networks
Subsample the original data set
Estimate the network with a fixed lambda (previously determined)
How often did I see an edge between 2 nodes relative to all subsamples
Sum up the variances of each edge Dn(lambda)
How does Stars select a model (lambda parmaters)
Given a curve of a summary statistic
monotonize
select model with beta param
Selection based on overall variability
The further to the right the curve gets the less edeges there are => little variability
How can I benchmark algorithms in the absence of a ground truth ?
Keyidea: Generate data that looks like real data based on a predefinced network structure
Zuletzt geändertvor 5 Monaten
