SPOTMisc (version 1.19.52)

kerasEvalPrediction: Evaluate keras prediction

Description

Evaluates prediction from keras model using several metrics based on training, validation and test data

Usage

kerasEvalPrediction(pred, testScore = c(NA, NA), specList, metrics, kerasConf)

Arguments

pred

prediction from keras predict

testScore

additional score values

specList

spec with target

metrics

keras metrics (history)

kerasConf

keras config

Examples

# \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){
library(tfdatasets)
library(keras)
target <- "age"
batch_size <- 32
prop <- 2/3
dfCensus <- getDataCensus(nobs=1000,
                          target = target)
data <- getGenericTrainValTestData(dfGeneric = dfCensus,
prop = prop)
specList <- genericDataPrep(data=data, batch_size = batch_size)
## spec test data has 334 elements:
str(specList$testGeneric$target)
## simulate test:
pred <- runif(length(specList$testGeneric$target))
kerasConf <- getKerasConf()
simpleModel <- getSimpleKerasModel(specList=specList,
                    kerasConf=kerasConf)
FLAGS <- list(epochs=16)
y <- kerasFit(model=simpleModel,
               specList = specList,
               FLAGS = FLAGS,
               kerasConf = kerasConf)
 simpeModel <- y$model
 history <- y$history
# evaluate on test data
pred <- predict(simpleModel, specList$testGeneric)
## in use keras evaluation (test error):
testScore <-
 keras::evaluate(simpleModel,
         tfdatasets::dataset_use_spec(dataset=specList$test_ds_generic,
         spec=specList$specGeneric_prep),
         verbose = kerasConf$verbose)
 kerasEvalPrediction(pred=pred,
                     testScore = testScore,
                    specList = specList,
                    metrics = history$metrics,
                    kerasConf = kerasConf
                    )
}
# }