R/calc_hlayers.R

In cranedroesch/panelNNET: Semiparametric panel data models using neural networks

calc_hlayers <- function(parlist, X = X, param = param, fe_var = fe_var, nlayers = nlayers, convolutional, activation){
  if (activation == 'tanh'){
    activ <- tanh
  }
  if (activation == 'logistic'){
    activ <- logistic
  }
  if (activation == 'relu'){
    activ <- relu
  }
  if (activation == 'lrelu'){
    activ <- lrelu
  }
  hlayers <- foreach(i = 1:length(nlayers)) %do% {
    vector('list', nlayers[i])
  }
  names(hlayers) <- paste0("p", 1:length(hlayers))
  for (p in 1:length(nlayers)){
    for (i in 1:(nlayers[p])){
      if (i == 1){D <- X[[p]]} else {D <- hlayers[[p]][[i-1]]}
      D <- as.matrix(cbind(1, D)) #add bias
      hlayers[[p]][[i]] <- activ(MatMult(D, parlist[[p]][[i]]))        
    }
    colnames(hlayers[[p]][[i]]) <- paste0("p_",p, '_nodes',1:ncol(hlayers[[p]][[i]]))
  }
  hlayers$param <- param
  if (!is.null(param)){
    colnames(hlayers$param) <- paste0('param',1:ncol(param))
  }
  if (is.null(fe_var)){#add intercept if no FEs
    hlayers$param <- cbind(rep(1, nrow(hlayers[[1]][[1]])), hlayers$param)
  }
  return(hlayers)
}