| .job.queue {StabPerf} | R Documentation |
Looks at the filesystem and determins how often feature-selectors and classifiers have to be sampled. Splits the workload into smaller jobs.
.job.queue(output, selectors, classifiers, predictors, cv=NULL)
output |
The directory were checkpoints are saved and looked for |
selectors |
A vector of expressions, each describing a selector (Take object data, return list of selected indices) |
classifiers |
A vector of expressions, each describing how to train a classifier (Take object data, return a model, that can be used by the corresponding predictor) |
predictors |
A vector of expressions, where the i-th expression can use a model model derived from the i-th classifier to predict the outcome of given expression values in data |
cv |
A list of vectors with indices, describing the samples to leave-out at the k-th fold in the cross-validation. NULL if no cross-validation is desired. |
Looks at the checkpoints written to output and determins what features from selectors and predictions from feature-selector/classifier-combinations have to be sampled. All sampling-step until sampling_steps * sampling_repeats are enqueued, in batches of sampling_repeats. If cv is given, jobs for cv-fold cross-validation are created.
jobs |
A list of jobs; can be passed to .compute.jobs, elements can be passed to compute.external |
compute.external, .compute.jobs
.job.queue(checkpointdir, selector.cmds, classifier.cmds, predictors.cmds, cv=NULL)