snnR (version 1.0)

snnR_extended: snnR_extended

Description

The snnR_extended function fits a sparse neural network for a GS model with additive and dominance efforts. It uses the sub-gradient and active-set methods to perform the optimization.

Usage

snnR_extended(x,y,z,nHidden_add,nHidden_dom,normalize=TRUE,verbose=TRUE, optimtol = 1e-5, 
 prgmtol = 1e-9, iteramax = 20, decsuff =1e-4,lambda)

Arguments

x

(numeric, \(n \times p\)) incidence matrix for additive effects.

y

(numeric, \(n\)) the response data-vector (NAs not allowed).

z

(numeric, \(n \times p\)) incidence matrix for dominance effects.

nHidden_add

(positive integer, \(1\times h\)) matrix, h indicates the number of hidden-layers and nHidden[1,h] indicates the neurons of the hth hidden-layer.

nHidden_dom

(positive integer, \(1\times h\)) matrix, h indicates the number of hidden-layers and nHidden[1,h] indicates the neurons of the hth hidden-layer.

normalize

logical, if TRUE normalizes output, the default value is FALSE.

verbose

logical, if TRUE prints detail history.

optimtol

numeric, a tiny number useful for checking convergenge of subgradients.

prgmtol

numeric, a tiny number useful for checking convergenge of parameters of NN.

iteramax

positive integer, maximum number of epochs(iterations) to train, default 20.

decsuff

numeric, a tiny number useful for checking change of loss function.

lambda

numeric, L1 norm lagrange multiplier.

Value

Mostly internal structure, but it is a list containing:

$wDNNs_add

A list containing weights and biases for additive effects.

$wDNNs_dom

A list containing weights and biases for dominance effects.

$inputwgts_add

A list containing input weights and biases for additive effects.

$outputwgts_add

A list containing output weights and biases for additive effects.

$hidewgts_add

A list containing hidden weights and biases for additive effects.

$inputwgts_dom

A list containing input weights and biases for dominance effects.

$outputwgts_dom

A list containing output weights and biases for dominance effects.

$hidewgts_dom

A list containing hidden weights and biases for dominance effects.

$Mse=Mse

The mean squared error between observed and predicted values.

$message

String that indicates the stopping criteria for the training process.

Examples

# NOT RUN {
###############################################################
#Example 1
  #Jersey dataset
   data(Jersey) 
  #Fit the model with Additive and Dominant effects
   y<-as.vector(pheno$yield_devMilk)
   X_test<-G[partitions==2,]
   X_train<-G[partitions!=2,]
   y_test<-y[partitions==2]
   y_train<-y[partitions!=2]
   Z_test<-D[partitions==2,]
   Z_train<-D[partitions!=2,]
   #Generate the structure of neural network   
   nHidden_add <- matrix(c(5,10,5),1,3)
   nHidden_dom <- matrix(c(5,15,5),1,3)
   # call function to train the sparse nerual network 
   network=snnR_extended(x=X_train,y=y_train,
   z=Z_train,nHidden_add=nHidden_add,nHidden_dom=nHidden_dom,iteramax =10,normalize=TRUE)
   # predictive results
   yhat= predict(network,X_test,Z_test)
   plot(y_test,yhat)

# }
# NOT RUN {
# }