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 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
|
.FilterEnsembleRegister = new.env() # nolint
#' Create an ensemble feature filter.
#'
#' Creates and registers custom ensemble feature filters. Implemented ensemble filters
#' can be listed with [listFilterEnsembleMethods]. Additional
#' documentation for the `fun` parameter specific to each filter can
#' be found in the description.
#'
#' @param name (`character(1)`)\cr
#' Identifier for the filter.
#' @param base.methods the base filter methods which the ensemble method
#' will use.
#' @param desc (`character(1)`)\cr
#' Short description of the filter.
#' @param fun (`function(task, nselect, ...`)\cr
#' Function which takes a task and returns a named numeric vector of scores,
#' one score for each feature of `task`.
#' Higher scores mean higher importance of the feature.
#' At least `nselect` features must be calculated, the remaining may be
#' set to `NA` or omitted, and thus will not be selected.
#' the original order will be restored if necessary.
#' @return Object of class \dQuote{FilterEnsemble}.
#' @export
#' @family filter
makeFilterEnsemble = function(name, base.methods, desc, fun) {
assertString(name)
assertString(desc)
assertFunction(fun, c("task", "base.methods"))
### calculate ensemble filter
obj = makeS3Obj("FilterEnsemble",
name = name,
desc = desc,
fun = fun
)
.FilterEnsembleRegister[[name]] = obj
obj
}
#' List ensemble filter methods.
#'
#' Returns a subset-able dataframe with filter information.
#'
#' @param desc (`logical(1)`)\cr
#' Provide more detailed information about filters.
#' Default is `TRUE`.
#' @return ([data.frame]).
#' @export
#' @family filter
listFilterEnsembleMethods = function(desc = TRUE) {
tag2df = function(tags, prefix = "") {
unique.tags = sort(unique(unlist(tags)))
res = asMatrixRows(lapply(tags, "%in%", x = unique.tags))
colnames(res) = stri_paste(prefix, unique.tags)
rownames(res) = NULL
as.data.frame(res)
}
assertFlag(desc)
filters = as.list(.FilterEnsembleRegister)
df = data.frame(
id = names(filters)
)
description = extractSubList(filters, "desc")
if (desc) {
df$desc = description
}
res = setRowNames(sortByCol(df, "id"), NULL)
addClasses(res, "FilterMethodsList")
return(res)
}
#' @export
print.FilterEnsembleMethodsList = function(x, len = 40, ...) {
if (!is.null(x$desc)) {
x$desc = clipString(x$desc, len = len)
}
NextMethod()
}
#' @export
print.FilterEnsemble = function(x, ...) {
catf("Filter: '%s'", x$name)
}
# E-min ----------------
#' Minimum ensemble filter. Takes the best minimum value across all base filter
#' methods for each feature.
#'
#' @rdname makeFilter
#' @name makeFilter
makeFilterEnsemble(
name = "E-min",
desc = "Minimum ensemble filter. Takes the best minimum value across all base filter methods for each feature.",
base.methods = NULL,
fun = function(task, base.methods, nselect, more.args) {
fval.all.ranked = rankBaseFilters(task = task, method = base.methods,
nselect = nselect, more.args = more.args)
### calculate ensemble filter
# group by "name" and summarize the minimum of "rank"
fval.ens = aggregate(fval.all.ranked$rank,
by = list(fval.all.ranked$name), FUN = min)
colnames(fval.ens) = c("name", "value")
# add columns "type" and "method"
fval.ens$type = fval.all.ranked$type[1:length(unique(fval.all.ranked$name))]
fval.ens$filter = "E-min"
# merge filters
fval.ens = mergeFilters(fval.all.ranked, fval.ens)
return(fval.ens)
}
)
# E-mean ----------------
#' Mean ensemble filter. Takes the mean across all base filter methods for each feature.
#'
#' @rdname makeFilter
#' @name makeFilter
makeFilterEnsemble(
name = "E-mean",
desc = "Mean ensemble filter. Takes the mean across all base filter methods for each feature.",
base.methods = NULL,
fun = function(task, base.methods, nselect, more.args) {
fval.all.ranked = rankBaseFilters(task = task, method = base.methods,
nselect = nselect, more.args = more.args)
### calculate ensemble filter
# group by "name" and summarize the minimum of "rank"
fval.ens = aggregate(fval.all.ranked$rank,
by = list(fval.all.ranked$name), FUN = mean)
colnames(fval.ens) = c("name", "value")
# add columns "type" and "method"
fval.ens$type = fval.all.ranked$type[1:length(unique(fval.all.ranked$name))]
fval.ens$filter = "E-mean"
# merge filters
fval.ens = mergeFilters(fval.all.ranked, fval.ens)
return(fval.ens)
}
)
# E-max ----------------
#' Maximum ensemble filter. Takes the best maximum value across all base filter
#' methods for each feature.
#'
#' @rdname makeFilter
#' @name makeFilter
makeFilterEnsemble(
name = "E-max",
desc = "Maximum ensemble filter. Takes the best maximum value across all base filter methods for each feature.",
base.methods = NULL,
fun = function(task, base.methods, nselect, more.args) {
fval.all.ranked = rankBaseFilters(task = task, method = base.methods,
nselect = nselect, more.args = more.args)
### calculate ensemble filter
# group by "name" and summarize the maximum of "rank"
fval.ens = aggregate(fval.all.ranked$rank,
by = list(fval.all.ranked$name), FUN = max)
colnames(fval.ens) = c("name", "value")
# add columns "type" and "method"
fval.ens$type = fval.all.ranked$type[1:length(unique(fval.all.ranked$name))]
fval.ens$filter = "E-max"
# merge filters
fval.ens = mergeFilters(fval.all.ranked, fval.ens)
return(fval.ens)
}
)
# E-median ----------------
#' Median ensemble filter. Takes the median across all base filter methods for
#' each feature.
#'
#' @rdname makeFilter
#' @name makeFilter
makeFilterEnsemble(
name = "E-median",
desc = "Median ensemble filter. Takes the median across all base filter methods for each feature.",
base.methods = NULL,
fun = function(task, base.methods, nselect, more.args) {
fval.all.ranked = rankBaseFilters(task = task, method = base.methods,
nselect = nselect, more.args = more.args)
### calculate ensemble filter
# group by "name" and summarize the median of "rank"
fval.ens = aggregate(fval.all.ranked$rank,
by = list(fval.all.ranked$name), FUN = median)
colnames(fval.ens) = c("name", "value")
# add columns "type" and "method"
fval.ens$type = fval.all.ranked$type[1:length(unique(fval.all.ranked$name))]
fval.ens$filter = "E-median"
# merge filters
fval.ens = mergeFilters(fval.all.ranked, fval.ens)
return(fval.ens)
}
)
# E-Borda ----------------
#' Borda ensemble filter. Takes the sum across all base filter methods for each
#' feature.
#'
#' @rdname makeFilter
#' @name makeFilter
makeFilterEnsemble(
name = "E-Borda",
desc = "Borda ensemble filter. Takes the sum across all base filter methods for each feature.",
base.methods = NULL,
fun = function(task, base.methods, nselect, more.args) {
if (length(unique(base.methods)) == 1L) {
stopf("Sampling without replacement is currently not supported for simple filter methods. Please use `makeDiscreteParam()` instead of `makeDiscreteVectorParam()`.")
}
fval.all.ranked = rankBaseFilters(task = task, method = base.methods,
nselect = nselect, more.args = more.args)
### calculate ensemble filter
# group by "name" and summarize the minimum of "rank"
fval.ens = aggregate(fval.all.ranked$rank,
by = list(fval.all.ranked$name), FUN = sum)
colnames(fval.ens) = c("name", "value")
# add columns "type" and "method"
fval.ens$type = fval.all.ranked$type[1:length(unique(fval.all.ranked$name))]
fval.ens$filter = "E-Borda"
# merge filters
fval.ens = mergeFilters(fval.all.ranked, fval.ens)
return(fval.ens)
}
)
# rank base filters -------------------------------------------------------
# helper fun to calculate and rank base filters for ensemble filters
rankBaseFilters = function(task, method = method,
nselect = nselect, more.args = more.args) {
# calculate base filters here
fval.calc = generateFilterValuesData(task, method = method,
nselect = nselect, more.args = more.args)
# rank base filters by method
value = NULL # due to NSE notes in R CMD check
fval.all.ranked = fval.calc$data[, rank := frank(value,
ties.method = "first"), by = filter]
setorderv(fval.all.ranked, c("filter", "rank"))
return(fval.all.ranked)
}
# merge base and ensemble filters ------------------------------------------------------
# helper fun to merge base and ensemble filters
mergeFilters = function(simple_filters, ensemble_filters) {
# merge ensemble and base filters
simple_filters$rank = NULL
all.filters = rbind(simple_filters, ensemble_filters)
return(all.filters)
}
|