workflowVariants: Generate (parameter) variants of a workflow

Description

The main goal of this function is to facilitate the generation of different variants of a workflow. The idea is to be able to supply several possible values for a set of parameters of the workflow, and then have the function to return a set of Workflow objects, each consisting of one of the different possible combinations of the variants. This function finds its use in the context of performance estimation experiments, where we may actually be interested in comparing different parameter settings for a workflow.

Usage

workflowVariants(wf,...,varsRootName,as.is=NULL)

Arguments

This is the string representing the name of the function of the base workflow from which variants should be generated. It can be ommited in the case of standard workflows (see Details section).

...

The function then accepts any number of named arguments, some of which may have vectors of values. These named arguments are supposed to be the names of parameters of the base workflow, and the parameters containing sets of values are the alternatives for the respective parameter that you want to consider in the variants generation (see examples below).

varsRootName

By default the names given to each variant will be formed by concatenating the base name of the workflow with the terminations: ".v1", ".v2", and so on. This parameter allows you to supply a different base name.

as.is

This is a vector of parameter names (defaults to NULL) that are not to be used as source for workflow variants. This is useful for workflows that have parameters that accept as "legal" values sets (e.g. a vector) and that we do not want the function workflowVariants to interprete as source values for generating different workflow variants but instead being use as they are given.

Value

The result of this function is a list of Workflow objects. Each of these objects represents one of the parameter variants of the workflow you have supplied.

References

Torgo, L. (2014) An Infra-Structure for Performance Estimation and Experimental Comparison of Predictive Models in R. arXiv:1412.0436 [cs.MS] http://arxiv.org/abs/1412.0436

Examples

Run this code

## Not run: 
# ## Generating several variants of the "svm" learner using different
# ## values of the parameter "cost", under the "umbrella" of a standard
# ## workflow (i.e. it assumes wf="standardWF")
# library(e1071)
# workflowVariants(learner="svm",cost=c(1,2,5,10))
# 
# ## variants of a user defined workflow (assumes that function "userWF"
# ## is defined and "understands" parameters par1 and par2)
# workflowVariants(wf="userWF",par1=c(0.1,0.4),par2=c(-10,10))
# 
# ## Variants of a standard time series workflows (it assumes that it is a
# ## time series workflow because of the use of the "type" parameter,
# ## otherwise you could make it explicit by adding wf="timeseriesWF").
# workflowVariants(learner="svm",type=c("slide","grow"),gamma=c(0.1,0.4))
# ## or equivalently
# workflowVariants(wf="timeseriesWF",learner="svm",type=c("slide","grow"),gamma=c(0.1,0.4))
# 
# ## allowing that one parameter is not considered for variants generation
# workflowVariants(wf="userWF",par1=c(0.1,0.4),par2=c(-10,10),as.is="par1")
# 
# ## nesting is also allowed!
# workflowVariants(wf="userWF",
#                  xpto=list(par1=c(0.1,0.4),d="k",par3=c(1,3)),
#                  par2=c(-10,10),
#                  as.is="par1")
# ## End(Not run)