R/makeSplit.R
In gero90000/MonitoringFeatures: Monitoring Features Generation
Defines functions
train_splitter
Documented in
train_splitter
#' Splitter Function
#'
#' @param split_data
#' @param valid_wanted
#' @param train_split
#' @param seed
#' @param ...
#'
#' @return
#' @export
#'
#' @examples
train_splitter = function(split_data, valid_wanted = FALSE, train_split, seed, ...){
base::set.seed(seed)
resls = list()
split_data = split_data
split_data_EAX = split_data[which(split_data$best_solver == "EAXrestart"), ]
split_data_LKH = split_data[which(split_data$best_solver == "LKHrestart"), ]
if(valid_wanted == FALSE){
# train test only
train_split = train_split
#=================================================================
# RUE
#=================================================================
# RUE.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "rue.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
rue.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
rue.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "rue.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# RUE.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "rue.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
rue.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
rue.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "rue.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# netgen.500
#=================================================================
# netgen.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
netgen.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
netgen.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "netgen.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# netgen.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
netgen.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
netgen.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "netgen.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# netgen.morphed.500
#=================================================================
# netgen.morphed.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
netgen.morphed.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
netgen.morphed.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# netgen.morphed.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
netgen.morphed.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
netgen.morphed.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# evol500.eaxsoph
#=================================================================
# evol500.eaxsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_sample_soph_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_soph_EAX, replace = F)
evol500.eaxsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
evol500.eaxsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph" &
split_data_EAX$ID %!in% samples_EAX), ]
# evol500.eaxsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_sample_soph_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_soph_LKH, replace = F)
evol500.eaxsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
evol500.eaxsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# evol500.lkhsoph // PROBLEM there are almost no lkh soph instancce where EAX+restart is the better solver
#=================================================================
# evol500.lkhsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_sample_soph_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_soph_EAX, replace = F) # with replacement
evol500.lkhsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
evol500.lkhsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph" &
split_data_EAX$ID %!in% samples_EAX), ]
# evol500.lkhsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_sample_soph_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_soph_LKH, replace = F) # with replacement
evol500.lkhsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
evol500.lkhsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# bind results
#=================================================================
TRAIN_train.500 = do.call("rbind", list(rue.500_sample_TRAIN_EAX,
rue.500_sample_TRAIN_LKH,
netgen.500_sample_TRAIN_EAX,
netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TRAIN_EAX,
netgen.morphed.500_sample_TRAIN_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX,
evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX,
evol500.lkhsoph.500_sample_TRAIN_LKH))
TEST_train.500 = do.call("rbind", list(rue.500_sample_TEST_EAX,
rue.500_sample_TEST_LKH,
netgen.500_sample_TEST_EAX,
netgen.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TEST_EAX,
evol500.lkhsoph.500_sample_TEST_LKH))
resls = list.append(resls,
TRAIN_train.500 = TRAIN_train.500,
TEST_train.500 = TEST_train.500
)
# garbage collect
rm(ID_sample_EAX, ID_sample_LKH, N_sample_other_LKH, N_sample_other_EAX,
N_sample_soph_EAX, N_sample_soph_LKH, samples_LKH, samples_EAX,
evol500.eaxsoph.500_sample_TEST_EAX, evol500.eaxsoph.500_sample_TEST_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX, evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TEST_EAX, evol500.lkhsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX, evol500.lkhsoph.500_sample_TRAIN_LKH,
netgen.500_sample_TEST_EAX, netgen.500_sample_TEST_LKH,
netgen.500_sample_TRAIN_EAX, netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TEST_EAX, netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TRAIN_EAX, netgen.morphed.500_sample_TRAIN_LKH,
rue.500_sample_TEST_EAX, rue.500_sample_TEST_LKH,
rue.500_sample_TRAIN_EAX, rue.500_sample_TRAIN_LKH,
train_split, split_data, split_data_EAX, split_data_LKH)
} else {
train_split = train_split
test_split = (1 - train_split) / 2
valid_split = test_split
# train valid and test
# actual sampling from each generator w.r.t. solver
#=================================================================
# RUE
#=================================================================
# RUE.EAX
# ID to sample from
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "rue.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
rue.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
rue.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
rue.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# RUE.LKH
# ID to sample from
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "rue.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
rue.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
rue.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
rue.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# netgen.500
#=================================================================
# netgen.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
netgen.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
netgen.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# netgen.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
netgen.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
netgen.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# netgen.morphed.500
#=================================================================
# netgen.morphed.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.morphed.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
netgen.morphed.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
netgen.morphed.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# netgen.morphed.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.morphed.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
netgen.morphed.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
netgen.morphed.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# evol500.eaxsoph
#=================================================================
# evol500.eaxsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.eaxsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
evol500.eaxsoph.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
evol500.eaxsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# evol500.eaxsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.eaxsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
evol500.eaxsoph.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
evol500.eaxsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# evol500.lkhsoph // PROBLEM there are almost no lkh soph instance where EAX+restart is the better solver
#=================================================================
# evol500.lkhsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.lkhsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
evol500.lkhsoph.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
evol500.lkhsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# evol500.lkhsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.lkhsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
evol500.lkhsoph.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
evol500.lkhsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# bind results
#=================================================================
TRAIN_train.500 = do.call("rbind", list(rue.500_sample_TRAIN_EAX,
rue.500_sample_TRAIN_LKH,
netgen.500_sample_TRAIN_EAX,
netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TRAIN_EAX,
netgen.morphed.500_sample_TRAIN_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX,
evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX,
evol500.lkhsoph.500_sample_TRAIN_LKH))
VALID_train.500 = do.call("rbind", list(rue.500_sample_VALID_EAX,
rue.500_sample_VALID_LKH,
netgen.500_sample_VALID_EAX,
netgen.500_sample_VALID_LKH,
netgen.morphed.500_sample_VALID_LKH,
netgen.morphed.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_VALID_EAX,
evol500.lkhsoph.500_sample_VALID_LKH))
TEST_train.500 = do.call("rbind", list(rue.500_sample_TEST_EAX,
rue.500_sample_TEST_LKH,
netgen.500_sample_TEST_EAX,
netgen.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TEST_EAX,
evol500.lkhsoph.500_sample_TEST_LKH))
resls = list.append(resls,
TRAIN_train.500 = TRAIN_train.500,
VALID_train.500 = VALID_train.500,
TEST_train.500 = TEST_train.500
)
}
rm(ID_sample_EAX, ID_sample_LKH, N_train, N_test, N_valid, samples_train, samples_test, samples_valid,
evol500.eaxsoph.500_sample_TEST_EAX, evol500.eaxsoph.500_sample_TEST_LKH,
evol500.eaxsoph.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_TRAIN_EAX, evol500.eaxsoph.500_sample_TRAIN_LKH, evol500.eaxsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_TEST_EAX, evol500.lkhsoph.500_sample_TEST_LKH, evol500.lkhsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX, evol500.lkhsoph.500_sample_VALID_EAX, evol500.lkhsoph.500_sample_TRAIN_LKH,
netgen.500_sample_TEST_EAX, netgen.500_sample_TEST_LKH, netgen.500_sample_VALID_LKH,
netgen.500_sample_TRAIN_EAX, netgen.500_sample_TRAIN_LKH, netgen.500_sample_VALID_EAX,
netgen.morphed.500_sample_TEST_EAX, netgen.morphed.500_sample_TEST_LKH, netgen.morphed.500_sample_VALID_LKH,
netgen.morphed.500_sample_TRAIN_EAX, netgen.morphed.500_sample_TRAIN_LKH,netgen.morphed.500_sample_VALID_EAX,
rue.500_sample_TEST_EAX, rue.500_sample_TEST_LKH, rue.500_sample_VALID_LKH,
rue.500_sample_TRAIN_EAX, rue.500_sample_TRAIN_LKH, rue.500_sample_VALID_EAX,
train_split, split_data, split_data_EAX, split_data_LKH)
return(resls)
}
gero90000/MonitoringFeatures documentation built on Dec. 17, 2020, 10:22 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions train_splitter
Documented in train_splitter
#' Splitter Function
#'
#' @param split_data
#' @param valid_wanted
#' @param train_split
#' @param seed
#' @param ...
#'
#' @return
#' @export
#'
#' @examples
train_splitter = function(split_data, valid_wanted = FALSE, train_split, seed, ...){
base::set.seed(seed)
resls = list()
split_data = split_data
split_data_EAX = split_data[which(split_data$best_solver == "EAXrestart"), ]
split_data_LKH = split_data[which(split_data$best_solver == "LKHrestart"), ]
if(valid_wanted == FALSE){
# train test only
train_split = train_split
#=================================================================
# RUE
#=================================================================
# RUE.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "rue.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
rue.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
rue.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "rue.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# RUE.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "rue.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
rue.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
rue.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "rue.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# netgen.500
#=================================================================
# netgen.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
netgen.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
netgen.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "netgen.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# netgen.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
netgen.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
netgen.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "netgen.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# netgen.morphed.500
#=================================================================
# netgen.morphed.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_sample_other_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_other_EAX, replace = F)
netgen.morphed.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
netgen.morphed.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500" &
split_data_EAX$ID %!in% samples_EAX), ]
# netgen.morphed.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_sample_other_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_other_LKH, replace = F)
netgen.morphed.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
netgen.morphed.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# evol500.eaxsoph
#=================================================================
# evol500.eaxsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_sample_soph_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_soph_EAX, replace = F)
evol500.eaxsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
evol500.eaxsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph" &
split_data_EAX$ID %!in% samples_EAX), ]
# evol500.eaxsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_sample_soph_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_soph_LKH, replace = F)
evol500.eaxsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
evol500.eaxsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# evol500.lkhsoph // PROBLEM there are almost no lkh soph instancce where EAX+restart is the better solver
#=================================================================
# evol500.lkhsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_sample_soph_EAX = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_EAX = sample(ID_sample_EAX, N_sample_soph_EAX, replace = F) # with replacement
evol500.lkhsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_EAX), ]
evol500.lkhsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph" &
split_data_EAX$ID %!in% samples_EAX), ]
# evol500.lkhsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_sample_soph_LKH = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
samples_LKH = sample(ID_sample_LKH, N_sample_soph_LKH, replace = F) # with replacement
evol500.lkhsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_LKH), ]
evol500.lkhsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph" &
split_data_LKH$ID %!in% samples_LKH), ]
#=================================================================
# bind results
#=================================================================
TRAIN_train.500 = do.call("rbind", list(rue.500_sample_TRAIN_EAX,
rue.500_sample_TRAIN_LKH,
netgen.500_sample_TRAIN_EAX,
netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TRAIN_EAX,
netgen.morphed.500_sample_TRAIN_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX,
evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX,
evol500.lkhsoph.500_sample_TRAIN_LKH))
TEST_train.500 = do.call("rbind", list(rue.500_sample_TEST_EAX,
rue.500_sample_TEST_LKH,
netgen.500_sample_TEST_EAX,
netgen.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TEST_EAX,
evol500.lkhsoph.500_sample_TEST_LKH))
resls = list.append(resls,
TRAIN_train.500 = TRAIN_train.500,
TEST_train.500 = TEST_train.500
)
# garbage collect
rm(ID_sample_EAX, ID_sample_LKH, N_sample_other_LKH, N_sample_other_EAX,
N_sample_soph_EAX, N_sample_soph_LKH, samples_LKH, samples_EAX,
evol500.eaxsoph.500_sample_TEST_EAX, evol500.eaxsoph.500_sample_TEST_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX, evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TEST_EAX, evol500.lkhsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX, evol500.lkhsoph.500_sample_TRAIN_LKH,
netgen.500_sample_TEST_EAX, netgen.500_sample_TEST_LKH,
netgen.500_sample_TRAIN_EAX, netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TEST_EAX, netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TRAIN_EAX, netgen.morphed.500_sample_TRAIN_LKH,
rue.500_sample_TEST_EAX, rue.500_sample_TEST_LKH,
rue.500_sample_TRAIN_EAX, rue.500_sample_TRAIN_LKH,
train_split, split_data, split_data_EAX, split_data_LKH)
} else {
train_split = train_split
test_split = (1 - train_split) / 2
valid_split = test_split
# train valid and test
# actual sampling from each generator w.r.t. solver
#=================================================================
# RUE
#=================================================================
# RUE.EAX
# ID to sample from
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "rue.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
rue.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
rue.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
rue.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# RUE.LKH
# ID to sample from
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "rue.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("rue.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
rue.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
rue.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
rue.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# netgen.500
#=================================================================
# netgen.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
netgen.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
netgen.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# netgen.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
netgen.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
netgen.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# netgen.morphed.500
#=================================================================
# netgen.morphed.500.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.morphed.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
netgen.morphed.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
netgen.morphed.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# netgen.morphed.500.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "netgen_morphed.500"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("netgen_morphed.500")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
netgen.morphed.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
netgen.morphed.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
netgen.morphed.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# evol500.eaxsoph
#=================================================================
# evol500.eaxsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.eaxsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
evol500.eaxsoph.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
evol500.eaxsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# evol500.eaxsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.eaxsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.eaxsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.eaxsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
evol500.eaxsoph.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
evol500.eaxsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# evol500.lkhsoph // PROBLEM there are almost no lkh soph instance where EAX+restart is the better solver
#=================================================================
# evol500.lkhsoph.EAX
ID_sample_EAX = (split_data_EAX[which(split_data_EAX$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_EAX[which(split_data_EAX$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_EAX) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_EAX) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_EAX, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_EAX, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_EAX, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.lkhsoph.500_sample_TRAIN_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_train), ]
evol500.lkhsoph.500_sample_VALID_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_valid), ]
evol500.lkhsoph.500_sample_TEST_EAX = split_data_EAX[which(split_data_EAX$ID %in% samples_test), ]
# evol500.lkhsoph.LKH
ID_sample_LKH = (split_data_LKH[which(split_data_LKH$generator == "evol500.lkhsoph"), "ID"]) %>% unique(.)
N_train = (((split_data_LKH[which(split_data_LKH$generator %in% c("evol500.lkhsoph")), "ID"]) %>%
unique(.) %>% length(.)) * train_split) %>% ceiling()
N_valid = ((length(ID_sample_LKH) - N_train)/2) %>% floor(.) # less for valid
N_test = ((length(ID_sample_LKH) - N_train)/2) %>% ceiling(.) # more for test
samples_train = sample(ID_sample_LKH, N_train, replace = F)
samples_valid = sample(setdiff(ID_sample_LKH, samples_train), N_valid, replace = F)
samples_test = sample(setdiff(ID_sample_LKH, samples_train) %>% setdiff(., samples_valid) , N_test, replace = F)
evol500.lkhsoph.500_sample_TRAIN_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_train), ]
evol500.lkhsoph.500_sample_VALID_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_valid), ]
evol500.lkhsoph.500_sample_TEST_LKH = split_data_LKH[which(split_data_LKH$ID %in% samples_test), ]
#=================================================================
# bind results
#=================================================================
TRAIN_train.500 = do.call("rbind", list(rue.500_sample_TRAIN_EAX,
rue.500_sample_TRAIN_LKH,
netgen.500_sample_TRAIN_EAX,
netgen.500_sample_TRAIN_LKH,
netgen.morphed.500_sample_TRAIN_EAX,
netgen.morphed.500_sample_TRAIN_LKH,
evol500.eaxsoph.500_sample_TRAIN_EAX,
evol500.eaxsoph.500_sample_TRAIN_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX,
evol500.lkhsoph.500_sample_TRAIN_LKH))
VALID_train.500 = do.call("rbind", list(rue.500_sample_VALID_EAX,
rue.500_sample_VALID_LKH,
netgen.500_sample_VALID_EAX,
netgen.500_sample_VALID_LKH,
netgen.morphed.500_sample_VALID_LKH,
netgen.morphed.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_VALID_EAX,
evol500.lkhsoph.500_sample_VALID_LKH))
TEST_train.500 = do.call("rbind", list(rue.500_sample_TEST_EAX,
rue.500_sample_TEST_LKH,
netgen.500_sample_TEST_EAX,
netgen.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_LKH,
netgen.morphed.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_EAX,
evol500.eaxsoph.500_sample_TEST_LKH,
evol500.lkhsoph.500_sample_TEST_EAX,
evol500.lkhsoph.500_sample_TEST_LKH))
resls = list.append(resls,
TRAIN_train.500 = TRAIN_train.500,
VALID_train.500 = VALID_train.500,
TEST_train.500 = TEST_train.500
)
}
rm(ID_sample_EAX, ID_sample_LKH, N_train, N_test, N_valid, samples_train, samples_test, samples_valid,
evol500.eaxsoph.500_sample_TEST_EAX, evol500.eaxsoph.500_sample_TEST_LKH,
evol500.eaxsoph.500_sample_VALID_EAX,
evol500.eaxsoph.500_sample_TRAIN_EAX, evol500.eaxsoph.500_sample_TRAIN_LKH, evol500.eaxsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_TEST_EAX, evol500.lkhsoph.500_sample_TEST_LKH, evol500.lkhsoph.500_sample_VALID_LKH,
evol500.lkhsoph.500_sample_TRAIN_EAX, evol500.lkhsoph.500_sample_VALID_EAX, evol500.lkhsoph.500_sample_TRAIN_LKH,
netgen.500_sample_TEST_EAX, netgen.500_sample_TEST_LKH, netgen.500_sample_VALID_LKH,
netgen.500_sample_TRAIN_EAX, netgen.500_sample_TRAIN_LKH, netgen.500_sample_VALID_EAX,
netgen.morphed.500_sample_TEST_EAX, netgen.morphed.500_sample_TEST_LKH, netgen.morphed.500_sample_VALID_LKH,
netgen.morphed.500_sample_TRAIN_EAX, netgen.morphed.500_sample_TRAIN_LKH,netgen.morphed.500_sample_VALID_EAX,
rue.500_sample_TEST_EAX, rue.500_sample_TEST_LKH, rue.500_sample_VALID_LKH,
rue.500_sample_TRAIN_EAX, rue.500_sample_TRAIN_LKH, rue.500_sample_VALID_EAX,
train_split, split_data, split_data_EAX, split_data_LKH)
return(resls)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.