autoencoder_gridsearch

##################
## FUNCTION 
##################

optAuto<-function(unit1,unit2,epo){
  # set model
  model <- keras_model_sequential()
  model %>%
    layer_dense(units = unit1, activation = "tanh", input_shape = ncol(M)) %>%
    #layer_dense(units = unit1, activation = "tanh")%>%
    layer_dense(units = unit2, activation = "tanh", name = "bottleneck") %>% 
    layer_dense(units = unit1, activation = "tanh") %>%
    layer_dense(units = ncol(M))
  
  # view model layers
  summary(model)
  # compile model
  model %>% compile(
    loss = "mean_squared_error", 
    optimizer = "adam"
  )
  
  # fit model
  model %>% fit(
    x = M, 
    y = M, 
    epochs = epo,
    verbose = 0
  )
  
  # evaluate the performance of the model
  mse.ae2 <- evaluate(model, M, M)
  return(model)
}

##################
## RUNNING GRID SEARCH
##################

# Create a data frame containing all combinations 
unit1 <- seq(40, 200, 20)
unit2 <- seq(7, 50, 7)
epo <- seq(8000, 12000, 1000)
hyper_grid <- expand.grid(unit1=unit1,unit2=unit2,epo=epo )

# Create an empty vector to store mse error values
mse_err <- c()

# Write a loop over the rows of hyper_grid to train the grid of models
for (i in 1:nrow(hyper_grid)) {
  
  # Train a Random Forest model
  model <- optAuto(unit1 = hyper_grid$unit1[i],
                   unit2 = hyper_grid$unit2[i],
                   epo = hyper_grid$epo[i])
  
  # Store MSE error for the model                      
  mse_err[i] <-model[[1]]
  print(mse_err[i])
  print(i)
}