Available options

library(CoOPLBM)

Options for the exploration procedure

All the following options concerns the exploration procedure of both fit_unsupervised_LBM and fit_unsupervised_CoOP_LBM functions, which is the part where the number of groups is changing in order to explore the best fit. The arguments must be in a list containing all the option.

Plots

To deactivate the plots use :

fit_unsupervised_CoOP_LBM(M,exploOptions = list(plot=F))

Verbosity

To deactivate all the prints, use:

fit_unsupervised_CoOP_LBM(M,exploOptions = list(verbosity=F))

Initialization type

There are several ways to initialize clustering during the exploration. The default initialization is hierarchical clustering. Spectral and K-means clustering are also available. If you want to change the initialization, use :

fit_unsupervised_CoOP_LBM(M,exploOptions = list(initMethod="spectral_clust"))
fit_unsupervised_CoOP_LBM(M,exploOptions = list(initMethod="kmeans_clust"))

Maximum Exploration

The algorithm keeps increasing the number of groups if the ICL increases. If the ICL decrease, the algorithm will keeps increasing the number of groups until it reaches a number of groups greater than 1.5 times the best matches. To change this value, use :

fit_unsupervised_CoOP_LBM(M,exploOptions = list(maxEplo=2))

Maximum Number of Groups

The algorithm keeps increasing the number of groups if the ICL increases, but will stop if the number of groups is greater than 10. To change that threshold, use :

fit_unsupervised_CoOP_LBM(M,exploOptions = list(maxGroups=15))

The threshold can’t be lower than 4.

Reinitialize the exploration

For a deeper exploration of the parameters, the exploration can be done backward and forward once the maximum of groups is found. This exploration leads to more precise result but is time consuming. Use :

fit_unsupervised_CoOP_LBM(M,exploOptions = list(reinitialize  = TRUE))

Options for the estimation procedure

All the following options concerns the estimation procedure, which is done for a number of fixed groups. These options are different for LBM and CoOP-LBM, but are available for both the unsupervised function and their supervised counterparts.

LBM

Maximum Iteration

To change the maximum number of iteration of the VEM algorithm, which default value is 50, use :

fit_unsupervised_LBM(M,estimOptions = list(maxIter  = 100))

Variational fixed point algorithm

The variational parameters are calculated with a fixed point algorithm, which default number of iteration is 3. To change the this value, use :

fit_unsupervised_LBM(M,estimOptions = list(fixPointIter  = 3))

Threshold of convergence

The estimation of the parameters stops if the relative distance between an estimation and the next is below a threshold equals to 1e-3. To change the threshold, use :

fit_unsupervised_LBM(M,estimOptions = list(threshold     = 1e-4))

CoOP-LBM

Maximum Iteration during burn in

To change the maximum number of iteration of the SEM algorithm during the burn in period, which default value is 50, use :

fit_unsupervised_CoOP_LBM(M,estimOptions = list(maxHeat  = 100))

Maximum Iteration after burn in

To change the maximum number of iteration of the SEM algorithm after the burn in period, which default value is 10000, use :

fit_unsupervised_CoOP_LBM(M,estimOptions = list(maxIter  = 100))

Fixed point algorithm

The sampling effort parameters are calculated with fixed point algorithm, which default number of iteration is 3. To change the this value, use :

fit_unsupervised_CoOP_LBM(M,estimOptions = list(fixPointIter  = 3))