3 #' Run the (core) agghoo procedure.
4 #' Arguments specify the list of models, their parameters and the
5 #' cross-validation settings, among others.
7 #' @param data Data frame or matrix containing the data in lines.
8 #' @param target The target values to predict. Generally a vector,
9 #' but possibly a matrix in the case of "soft classification".
10 #' @param task "classification" or "regression". Default:
11 #' regression if target is numerical, classification otherwise.
12 #' @param gmodel A "generic model", which is a function returning a predict
13 #' function (taking X as only argument) from the tuple
14 #' (dataHO, targetHO, param), where 'HO' stands for 'Hold-Out',
15 #' referring to cross-validation. Cross-validation is run on an array
16 #' of 'param's. See params argument. Default: see R6::Model.
17 #' @param params A list of parameters. Often, one list cell is just a
18 #' numerical value, but in general it could be of any type.
19 #' Default: see R6::Model.
20 #' @param loss A function assessing the error of a prediction.
21 #' Arguments are y1 and y2 (comparing a prediction to known values).
22 #' loss(y1, y2) --> real number (error). Default: see R6::AgghooCV.
25 #' An R6::AgghooCV object o. Then, call o$fit() and finally o$predict(newData)
29 #' a_reg <- agghoo(iris[,-c(2,5)], iris[,2])
31 #' pr <- a_reg$predict(iris[,-c(2,5)] + rnorm(450, sd=0.1))
33 #' a_cla <- agghoo(iris[,-5], iris[,5])
35 #' pc <- a_cla$predict(iris[,-5] + rnorm(600, sd=0.1))
37 #' @seealso Function \code{\link{compareTo}}
40 #' Guillaume Maillard, Sylvain Arlot, Matthieu Lerasle. "Aggregated hold-out".
41 #' Journal of Machine Learning Research 22(20):1--55, 2021.
44 agghoo <- function(data, target, task = NULL, gmodel = NULL, params = NULL, loss = NULL) {
46 if (!is.data.frame(data) && !is.matrix(data))
47 stop("data: data.frame or matrix")
48 if (is.data.frame(target) || is.matrix(target)) {
49 if (nrow(target) != nrow(data) || ncol(target) == 1)
50 stop("target probability matrix does not match data size")
52 else if (!is.numeric(target) && !is.factor(target) && !is.character(target))
53 stop("target: numeric, factor or character vector")
55 task = match.arg(task, c("classification", "regression"))
56 if (is.character(gmodel))
57 gmodel <- match.arg(gmodel, c("knn", "ppr", "rf", "tree"))
58 else if (!is.null(gmodel) && !is.function(gmodel))
59 # No further checks here: fingers crossed :)
60 stop("gmodel: function(dataHO, targetHO, param) --> function(X) --> y")
61 if (is.numeric(params) || is.character(params))
62 params <- as.list(params)
63 if (!is.list(params) && !is.null(params))
64 stop("params: numerical, character, or list (passed to model)")
65 if (is.function(gmodel) && !is.list(params))
66 stop("params must be provided when using a custom model")
67 if (is.list(params) && is.null(gmodel))
68 stop("model (or family) must be provided when using custom params")
69 if (!is.null(loss) && !is.function(loss))
70 # No more checks here as well... TODO:?
71 stop("loss: function(y1, y2) --> Real")
74 if (is.numeric(target))
77 task = "classification"
79 # Build Model object (= list of parameterized models)
80 model <- Model$new(data, target, task, gmodel, params)
81 # Return AgghooCV object, to run and predict
82 AgghooCV$new(data, target, task, model, loss)