estim_delta: Estimates delta for Iterated Stable Autoencoder

Description Usage Arguments Details Value See Also Examples

Description

This function uses cross-validation to estimate delta for the Iterated Stable Autoencoder when considering Binomial noise. delta is the probability of deletion of each cell of the data matrix

Usage

1
2
3
estim_delta(X, delta = seq(0.1, 0.9, length.out = 9), nbsim = 10,
  noise = "Binomial", transformation = c("None", "CA"), pNA = 0.1,
  maxiter = 1000, threshold = 1e-08)

Arguments

X

a data frame or a matrix with count

delta

vector, a sequence of values for the probability of deletion of each cell of the data matrix

nbsim

number of times that pNA values are inserted and predicted in the data

noise

noise model assumed for the data. By default and only available "Binomial"

transformation

estimates a transformation of the original matrix; currently, only correspondence analysis CA is available

pNA

percentage of missing values added in the data set

maxiter

integer, maximum number of iterations of the iterative imputation algorithm

threshold

for assessing convergence of the iterative imputation algorithm (difference between two successive iterations)

Details

For each value delta, repeated learning cross-validation consists in inserting pNA percentage of missing values in the data set and predicting them with the Iterative Stable Autoencoder. More precisely, the prediction is obtained using the iterative imputation algorithm (imputecount) which alternates steps of imputation of the missing entries and estimation of the low-rank signal. This process is repeated nbsim times for all the deltas. The mean squared error of prediction is kept for each simulation and value of delta. The value of delta leading to the smallest MSEP on average over the simulations is given.

Value

msep, matrix with the MSEP obtained for each simulation and each value of delta

delta, value giving in average the smallest MSEP over the nbsim simulations

See Also

imputecount

ISA

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
 # A regularized Correspondence Analysis 
 ## Not run: library(FactoMineR)
 perfume <-  read.table("http://factominer.free.fr/docs/perfume.txt",header=TRUE,
 sep="\t",row.names=1)
 rownames(perfume)[4] <- "Cinema"
 
 isa.delt <- estim_delta(perfume, nbsim = 10, transformation = "CA")
 
 isa.ca <- ISA(perfume, delta = isa.delt$delta, noise = "Binomial", transformation = "CA")
 rownames(isa.ca$mu.hat) <- rownames(perfume)
 colnames(isa.ca$mu.hat) <- colnames(perfume)
 res.isa.ca <- CA(isa.ca$mu.hat, graph = FALSE)
 plot(res.isa.ca, title = "Regularized CA", cex = 0.6, selectCol = "contrib 20")
## End(Not run)

denoiseR documentation built on May 1, 2019, 10:45 p.m.