-
Notifications
You must be signed in to change notification settings - Fork 26
/
finalize.R
283 lines (248 loc) · 7.09 KB
/
finalize.R
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
#' Functions to finalize data-specific parameter ranges
#'
#' These functions take a parameter object and modify the unknown parts of
#' `ranges` based on a data set and simple heuristics.
#'
#' @param object A `param` object or a list of `param` objects.
#'
#' @param x The predictor data. In some cases (see below) this should only
#' include numeric data.
#'
#' @param force A single logical that indicates that even if the parameter
#' object is complete, should it update the ranges anyway?
#'
#' @param log_vals A logical: should the ranges be set on the log10 scale?
#'
#' @param ... Other arguments to pass to the underlying parameter
#' finalizer functions. For example, for `get_rbf_range()`, the dots are passed
#' along to [kernlab::sigest()].
#'
#' @param frac A double for the fraction of the data to be used for the upper
#' bound. For `get_n_frac_range()` and `get_batch_sizes()`, a vector of two
#' fractional values are required.
#'
#' @param seed An integer to control the randomness of the calculations.
#'
#' @return
#'
#' An updated `param` object or a list of updated `param` objects depending
#' on what is provided in `object`.
#'
#' @details
#'
#' `finalize()` runs the embedded finalizer function contained in the `param`
#' object (`object$finalize`) and returns the updated version. The finalization
#' function is one of the `get_*()` helpers.
#'
#' The `get_*()` helper functions are designed to be used with the pipe
#' and update the parameter object in-place.
#'
#' `get_p()` and `get_log_p()` set the upper value of the range to be
#' the number of columns in the data (on the natural and
#' log10 scale, respectively).
#'
#' `get_n()` and `get_n_frac()` set the upper value to be the number of
#' rows in the data or a fraction of the total number of rows.
#'
#' `get_rbf_range()` sets both bounds based on the heuristic defined in
#' [kernlab::sigest()]. It requires that all columns in `x` be numeric.
#'
#' @examplesIf interactive() || identical(Sys.getenv("IN_PKGDOWN"), "true")
#' library(dplyr)
#' car_pred <- select(mtcars, -mpg)
#'
#' # Needs an upper bound
#' mtry()
#' finalize(mtry(), car_pred)
#'
#' # Nothing to do here since no unknowns
#' penalty()
#' finalize(penalty(), car_pred)
#'
#' library(kernlab)
#' library(tibble)
#' library(purrr)
#'
#' params <-
#' tribble(
#' ~parameter, ~object,
#' "mtry", mtry(),
#' "num_terms", num_terms(),
#' "rbf_sigma", rbf_sigma()
#' )
#' params
#'
#' # Note that `rbf_sigma()` has a default range that does not need to be
#' # finalized but will be changed if used in the function:
#' complete_params <-
#' params %>%
#' mutate(object = map(object, finalize, car_pred))
#' complete_params
#'
#' params %>%
#' dplyr::filter(parameter == "rbf_sigma") %>%
#' pull(object)
#' complete_params %>%
#' dplyr::filter(parameter == "rbf_sigma") %>%
#' pull(object)
#'
#' @export
finalize <- function(object, ...) {
UseMethod("finalize")
}
#' @export
#' @rdname finalize
finalize.list <- function(object, x, force = TRUE, ...) {
map(object, finalize, x, force, ...)
}
#' @export
#' @rdname finalize
finalize.param <- function(object, x, force = TRUE, ...) {
if (is.null(object$finalize)) {
return(object)
}
if (!has_unknowns(object) & !force) {
return(object)
}
object$finalize(object, x = x, ...)
}
safe_finalize <- function(object, x, force = TRUE, ...) {
if (all(is.na(object))) {
res <- NA
} else {
res <- finalize(object, x, force = TRUE, ...)
}
res
}
#' @export
#' @rdname finalize
finalize.parameters <- function(object, x, force = TRUE, ...) {
object$object <- map(object$object, safe_finalize, x, force, ...)
object
}
# These two finalize methods are for cases when a tuning parameter has no
# parameter object or isn't listed in the tunable method.
#' @export
#' @rdname finalize
finalize.logical <- function(object, x, force = TRUE, ...) {
object
}
#' @export
#' @rdname finalize
finalize.default <- function(object, x, force = TRUE, ...) {
if (all(is.na(object))) {
return(object)
} else {
cls <- paste0("'", class(x), "'", collapse = ", ")
rlang::abort(paste0("Cannot finalize an object with class(es): ", cls))
}
object
}
#' @export
#' @rdname finalize
get_p <- function(object, x, log_vals = FALSE, ...) {
check_param(object)
rngs <- range_get(object, original = FALSE)
if (!is_unknown(rngs$upper)) {
return(object)
}
x_dims <- dim(x)
if (is.null(x_dims)) {
rlang::abort("Cannot determine number of columns. Is `x` a 2D data object?")
}
if (log_vals) {
rngs[2] <- log10(x_dims[2])
} else {
rngs[2] <- x_dims[2]
}
if (object$type == "integer" & is.null(object$trans)) {
rngs <- as.integer(rngs)
}
range_set(object, rngs)
}
#' @export
#' @rdname finalize
get_log_p <- function(object, x, ...) {
get_p(object, x, log_vals = TRUE, ...)
}
#' @export
#' @rdname finalize
get_n_frac <- function(object, x, log_vals = FALSE, frac = 1/3, ...) {
check_param(object)
rngs <- range_get(object, original = FALSE)
if (!is_unknown(rngs$upper)) {
return(object)
}
x_dims <- dim(x)
if (is.null(x_dims)) {
rlang::abort("Cannot determine number of columns. Is `x` a 2D data object?")
}
n_frac <- floor(x_dims[1] * frac)
if (log_vals) {
rngs[2] <- log10(n_frac)
} else {
rngs[2] <- n_frac
}
if (object$type == "integer" & is.null(object$trans) & !log_vals) {
rngs <- as.integer(rngs)
}
range_set(object, rngs)
}
#' @export
#' @rdname finalize
get_n_frac_range <- function(object, x, log_vals = FALSE, frac = c(1/10, 5/10), ...) {
rngs <- range_get(object, original = FALSE)
if (!is_unknown(rngs$upper)) {
return(object)
}
x_dims <- dim(x)
if (is.null(x_dims)) {
rlang::abort("Cannot determine number of columns. Is `x` a 2D data object?")
}
n_frac <- sort(floor(x_dims[1] * frac))
if (log_vals) {
rngs <- log10(n_frac)
} else {
rngs <- n_frac
}
if (object$type == "integer" & is.null(object$trans) & !log_vals) {
rngs <- as.integer(rngs)
}
range_set(object, rngs)
}
#' @export
#' @rdname finalize
get_n <- function(object, x, log_vals = FALSE, ...) {
get_n_frac(object, x, log_vals, frac = 1, ...)
}
#' @export
#' @rdname finalize
get_rbf_range <- function(object, x, seed = sample.int(10^5, 1), ...) {
rlang::check_installed("kernlab")
suppressPackageStartupMessages(requireNamespace("kernlab", quietly = TRUE))
x_mat <- as.matrix(x)
if (!is.numeric(x_mat)) {
rlang::abort("The matrix version of the initialization data is not numeric.")
}
with_seed(seed, rng <- kernlab::sigest(x_mat, ...)[-2])
rng <- log10(rng)
range_set(object, rng)
}
#' @export
#' @rdname finalize
get_batch_sizes <- function(object, x, frac = c(1/10, 1/3), ...) {
rngs <- range_get(object, original = FALSE)
if (!is_unknown(rngs$lower) & !is_unknown(rngs$upper)) {
return(object)
}
x_dims <- dim(x)
if (is.null(x_dims)) {
rlang::abort("Cannot determine number of columns. Is `x` a 2D data object?")
}
n_frac <- sort(floor(x_dims[1] * frac))
n_frac <- log2(n_frac)
if (object$type == "integer" & is.null(object$trans)) {
n_frac <- as.integer(n_frac)
}
range_set(object, n_frac)
}