Noisy Stochastic Block Mode: Graph Inference by Multiple Testing
compute conditional l-values in the noisy stochastic block model
main function of VEM algorithm with fixed number of SBM blocks
main function of VEM algorithm for fixed number of latent blocks in pa...
evaluate the density in the current model
M-step
plot the data matrix, the inferred graph and/or the true binary graph
plot ICL curve
auxiliary function for the computation of q-values
compute q-values in the noisy stochastic block model
simulation of a graph according the noisy stochastic block model
spectral clustering with absolute values
Create new initial values by merging pairs of groups of provided tau
Create new values of tau by splitting groups of provided tau
Compute one iteration to solve the fixed point equation in the VE-step
Perform one iteration of the Newton-Raphson to compute the MLE of the ...
VE-step
split group q of provided tau randomly into two into
Evalute the adjusted Rand index
convert a clustering into a 0-1-matrix
transform a pair of block identifiers (q,l) into an identifying intege...
takes a scalar indice of a group pair (q,l) and returns the values q a...
transform a pair of nodes (i,j) into an identifying integer
corrects values of the variational parameters tau that are too close t...
compute the MLE in the Gamma model using the Newton-Raphson method
VEM algorithm to adjust the noisy stochastic block model to an observe...
optimal number of SBM blocks
compute rho associated with given values of w, nu0 and nu
Evaluate tau_q*tau_l in the noisy stochastic block model
new graph inference procedure
computation of the Integrated Classification Likelihood criterion
compute a list of initial points for the VEM algorithm
Construct initial values with Q groups by meging groups of a solution ...
Construct initial values with Q groups by splitting groups of a soluti...
compute initial values of rho
compute intial values for tau
evaluate the objective in the Gamma model
evaluation of the objective in the Gauss model
returns a list of all possible node pairs (i,j)
Variational Expectation-Maximization algorithm to fit the noisy stochastic block model to an observed dense graph and to perform a node clustering. Moreover, a graph inference procedure to recover the underlying binary graph. This procedure comes with a control of the false discovery rate. The method is described in the article "Powerful graph inference with false discovery rate control" by T. Rebafka, E. Roquain, F. Villers (2020) <arXiv:1907.10176>.