originalThresholdout: Original Thresholdout algorithm

Description Usage Arguments Value See Also Examples

View source: R/classification.R

Description

Original Thresholdout with Dwork’s linear classifier (Dwork, et al., 2015)

Usage

1
2
3
4
5
originalThresholdout(train.ds = NULL, holdout.ds = NULL,
  validation.ds = NULL, label = "class", is.simulated = TRUE,
  update.freq = 50, pec.file = NULL, threshold = 4/sqrt(nrow(holdout.ds)),
  tolerance = 1/sqrt(nrow(holdout.ds)), signal.names = NULL,
  save.file = NULL, verbose = FALSE)

Arguments

train.ds

A data frame with training data and outcome labels

holdout.ds

A data frame with holdout data and outcome labels

validation.ds

A data frame with validation data and outcome labels

label

A character vector of the outcome variable column name

is.simulated

Is the data simulated (or real?)

update.freq

A integer for the number of steps before update

pec.file

A character vector filename of privateEC results

threshold

A numeric, default 4 / sqrt(n) suggested in the thresholdout’s supplementary material (Dwork, et al.,2015)

tolerance

A numeric, default 1 / sqrt(n) suggested in the thresholdout’s supplementary material (Dwork, et al.,2015)

signal.names

A character vector of signal names in simulated data

save.file

A character vector for results filename or NULL to skip

verbose

A flag indicating whether verbose output be sent to stdout

Value

A list containing:

algo.acc

data frame of results, a row for each update

ggplot.data

melted results data frame for plotting

correct

number of variables detected correctly in each data set

elapsed

total elapsed time

See Also

Other classification: epistasisRank, getImportanceScores, privateEC, privateRF, standardRF, xgboostRF

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
num.samples <- 100
num.variables <- 100
pct.signals <- 0.1
temp.pec.file <- tempfile(pattern = "pEc_temp", tmpdir = tempdir())
sim.data <- createSimulation(num.variables = num.variables,
                             num.samples = num.samples,
                             sim.type = "mainEffect",
                             pct.train = 1 / 3,
                             pct.holdout = 1 / 3,
                             pct.validation = 1 / 3,
                             verbose = FALSE)
pec.results <- privateEC(train.ds = sim.data$train,
                         holdout.ds = sim.data$holdout,
                         validation.ds = sim.data$validation,
                         label = sim.data$label,
                         is.simulated = TRUE,
                         signal.names = sim.data$signal.names,
                         save.file = temp.pec.file,
                         verbose = FALSE)
por.results <- originalThresholdout(train.ds = sim.data$train,
                                    holdout.ds = sim.data$holdout,
                                    validation.ds = sim.data$validation,
                                    label = sim.data$label,
                                    is.simulated = TRUE,
                                    signal.names = sim.data$signal.names,
                                    pec.file = temp.pec.file,
                                    verbose = FALSE)
file.remove(temp.pec.file)

hexhead/privateEC documentation built on July 20, 2018, 12:30 p.m.