ActiveLearningOptimizer

See Javadoc for complete documentation of this class.

Usage: <Random fraction> <Worst fraction>
Random fraction: Fraction of entries each generation that will be randomly chosen
Worst fraction: Fraction of entries to select from the entries with least-reliable predictions

Available Operations

These commands can be used to perform a variety of tasks, ranging from defining important settings about the object to actually using it.

run – Start optimizer
Will run until max iterations is reached, or oracle is unable to process all new candidates.

set gensize <number – Set number of new candidates per iteration
number: Desired number of new entries per generation

set initial $<dataset> – Define initial population
dataset: Dataset containing entries to use as initial population

set maxiter <number> – Set the number of iterations algorithm will perform
number: Desired maximum number of iterations

set model $<model> – Define group of models to train during optimization
model: Model to iteratively train and use predictively. Must implement CompositeRegression.

set objective <min|max> <method> [<options...>] – Set the objective function
min|max: Whether to minimize or maximize the objective function.
method: Desired objective function. Name of an BaseEntryRanker class. ("?" for options)
options...: Options for objective function

set oracle <method> [<options...>] – Define method used to calculate properties of selected candidates
method: How to calculate properties of an entry. Name of a BaseOracle ("?" for options)

set search $<dataset> – Define search space
dataset: Dataset containing any and all entries optimization algorithm is allowed to evaluate

stats ntop <number> – Control number of top entries to detect when evaluating algorithm performance
number: Number of globally best-performing entries to check for after each generation

stats success <include|exclude> <filter method> [<filter options...>] – Define a filter used to define when an entry is a "success"
include|exclude: Whether to include entries that pass the filter as a success
filter method: Name of a dataset filter ("?" for options)
filter options...: Any options for that filter

write generations <directory> <format> – Write out the data from each generation
directory: Directory in which to save datasets
format: Format in which to write datasets

Available Print Commands

These commands are run by calling "print <variable name> <command> [<options>]". Any output from that command will be printed to standard output.

stats – Print statistics about optimization run