Fit sparse PCA on a grid of alpha

Fit sparse PCA on a grid of alpha

FitSparsePCA(M, r = 1, min.size = 1, alphaGrid = 10^(seq(-4, 0, by = 0.1)))

Arguments

M

Gaussian proxy for the original dataset. Possibly obtained using PLNmodels.
r

Number of missing actors.
min.size

Minimal number of neighbors for each missing actor.
alphaGrid

Grid for parsity controlling parameter.

Value

  • sPcaOpt: optimal spca object.

  • alphaOpt: best alpha value among the provided grid.

  • loglik: vector of log likelihood obtained for each value of alpha.

  • bic: vecotr of BIC values.

  • cliques: optimal clique of neighbors.

Examples

data=generate_missing_data(n=100,p=10,r=1,type="scale-free", plot=TRUE)

TrueClique=data$TC
PLNfit=norm_PLN(data$Y)
findclique=FitSparsePCA(PLNfit$MO,r=1)
initClique=findclique$cliques
TrueClique

#> [[1]]
#> [1] 1 4 7 8 9
#> 
initClique

#> [[1]]
#> [1]  2  4  5  6  7  8  9 10
#>