funKerasGeneric: funKerasGeneric

Description

Hyperparameter Tuning: Generic Classification Objective Function.

Usage

funKerasGeneric(x, kerasConf = NULL, specList = NULL)

Value

1-column matrix with resulting function values (test loss)

Arguments

x: matrix of hyperparameter values to evaluate with the function. Rows for points and columns for dimension.
kerasConf: List of additional parameters passed to keras as described in getKerasConf. Default: NULL.
specList: prepared data. See genericDataPrep.

Details

Trains a simple deep NN on arbitrary data sets. Provides a template that can be used for other networks as well. Standard Code from https://tensorflow.rstudio.com/ Modified by T. Bartz-Beielstein (tbb@bartzundbartz.de)

Note: The WARNING "tensorflow:Layers in a Sequential model should only have a single input tensor. Consider rewriting this model with the Functional API" can be safely ignored: in general, Keras encourages its users to use functional models for multi-input layers, but there is nothing wrong with doing so. See: https://github.com/tensorflow/recommenders/issues/188.

Examples

Run this code

# \donttest{
### These examples require an activated Python environment as described in
### Bartz-Beielstein, T., Rehbach, F., Sen, A., and Zaefferer, M.:
### Surrogate Model Based Hyperparameter Tuning for Deep Learning with SPOT,
### June 2021. http://arxiv.org/abs/2105.14625.
PYTHON_RETICULATE <- FALSE
if(PYTHON_RETICULATE){

## data preparation
target <- "age"
batch_size <- 32
prop <- 2/3
dfGeneric <- getDataCensus(target = target, nobs = 1000)
data <- getGenericTrainValTestData(dfGeneric = dfGeneric, prop = prop)
specList <- genericDataPrep(data=data, batch_size = batch_size)

## model configuration:
cfg <-  getModelConf(list(model="dl"))
x <- matrix(cfg$default, nrow=1)
transformFun <- cfg$transformations
types <- cfg$type
lower <- cfg$lower
upper <- cfg$upper

kerasConf <- getKerasConf()

### First example: simple function call:
message("objectiveFunctionEvaluation(): x before transformX().")
print(x)
if (length(transformFun) > 0) {  x <- transformX(xNat=x, fn=transformFun)}
message("objectiveFunctionEvaluation(): x after transformX().")
print(x)
funKerasGeneric(x, kerasConf = kerasConf, specList = specList)

### Second example: evaluation of several (three) hyperparameter settings:
xxx <- rbind(x,x,x)
funKerasGeneric(xxx, kerasConf = kerasConf, specList)

### Third example: spot call with extended verbosity:
res <- spot(x = NULL,
            fun = funKerasGeneric,
            lower = lower,
            upper = upper,
            control = list(funEvals=50,
                         handleNAsMethod = handleNAsMean,
                         noise = TRUE,
                         types = types,
                         plots = TRUE,
                         progress = TRUE,
                         seedFun = 1,
                         seedSPOT = 1,
                         transformFun=transformFun),
                         kerasConf = kerasConf,
                         specList = specList)
  }
# }