Define Problems for Experiments

Problems may consist of up to two parts: A static, immutable part (data in addProblem) and a dynamic, stochastic part (fun in addProblem). For example, for statistical learning problems a data frame would be the static problem part while a resampling function would be the stochastic part which creates problem instance. This instance is then typically passed to a learning algorithm like a wrapper around a statistical model (fun in addAlgorithm).

This function serialize all components to the file system and registers the problem in the ExperimentRegistry.

removeProblem removes all jobs from the registry which depend on the specific problem. reg$problems holds the IDs of already defined problems.

addProblem(
  name,
  data = NULL,
  fun = NULL,
  seed = NULL,
  cache = FALSE,
  reg = getDefaultRegistry()
)

removeProblems(name, reg = getDefaultRegistry())

Arguments

name	[`character(1)`] Unique identifier for the problem.
data	[`ANY`] Static problem part. Default is `NULL`.
fun	[`function`] The function defining the stochastic problem part. The static part is passed to this function with name “data” and the `Job`/`Experiment` is passed as “job”. Therefore, your function must have the formal arguments “job” and “data” (or dots `...`). If you do not provide a function, it defaults to a function which just returns the data part.
seed	[`integer(1)`] Start seed for this problem. This allows the “synchronization” of a stochastic problem across algorithms, so that different algorithms are evaluated on the same stochastic instance. If the problem seed is defined, the seeding mechanism works as follows: (1) Before the dynamic part of a problem is instantiated, the seed of the problem + [replication number] - 1 is set, i.e. the first replication uses the problem seed. (2) The stochastic part of the problem is instantiated. (3) From now on the usual experiment seed of the registry is used, see `ExperimentRegistry`. If `seed` is set to `NULL` (default), the job seed is used to instantiate the problem and different algorithms see different stochastic instances of the same problem.
cache	[`logical(1)`] If `TRUE` and `seed` is set, problem instances will be cached on the file system. This assumes that each problem instance is deterministic for each combination of hyperparameter setting and each replication number. This feature is experimental.
reg	[`ExperimentRegistry`] Registry. If not explicitly passed, uses the last created registry.

Value

[Problem]. Object of class “Problem” (invisibly).

Examples

 batchtools:::example_push_temp(1) 
tmp = makeExperimentRegistry(file.dir = NA, make.default = FALSE)
#> No readable configuration file found
#> Created registry in '/tmp/batchtools-example/reg' using cluster functions 'Interactive'
addProblem("p1", fun = function(job, data) data, reg = tmp)
#> Adding problem 'p1'
addProblem("p2", fun = function(job, data) job, reg = tmp)
#> Adding problem 'p2'
addAlgorithm("a1", fun = function(job, data, instance) instance, reg = tmp)
#> Adding algorithm 'a1'
addExperiments(repls = 2, reg = tmp)
#> Adding 2 experiments ('p1'[1] x 'a1'[1] x repls[2]) ...
#> Adding 2 experiments ('p2'[1] x 'a1'[1] x repls[2]) ...

# List problems, algorithms and job parameters:
tmp$problems
#> [1] "p1" "p2"
tmp$algorithms
#> [1] "a1"
getJobPars(reg = tmp)
#>    job.id problem prob.pars algorithm algo.pars
#> 1:      1      p1 <list[0]>        a1 <list[0]>
#> 2:      2      p1 <list[0]>        a1 <list[0]>
#> 3:      3      p2 <list[0]>        a1 <list[0]>
#> 4:      4      p2 <list[0]>        a1 <list[0]>

# Remove one problem
removeProblems("p1", reg = tmp)
#> Removing Problem 'p1' and 2 corresponding jobs ...

# List problems and algorithms:
tmp$problems
#> [1] "p2"
tmp$algorithms
#> [1] "a1"
getJobPars(reg = tmp)
#>    job.id problem prob.pars algorithm algo.pars
#> 1:      3      p2 <list[0]>        a1 <list[0]>
#> 2:      4      p2 <list[0]>        a1 <list[0]>

Arguments

Value

See also

Examples