Base class for objects that propose the next active-learning batch.
Proposers receive an ALContext from OptimizerAL. They should use the context's accessor methods instead of touching surrogate and acquisition registries directly, because those methods implement lazy update and clone-on-write rules.
Inheriting classes should implement the private$.propose() method.
It gets the context and the number of points to propose, should return a
data.table of the proposed points.
Creates a new proposer.
Proposes up to n points for the next evaluation batch.
Calculates the scalar score to be maximized by the proposer.
Returns the indices of the top n candidates based on the utility score.
Arguments
- id
(
character(1))
Identifier.- param_set
(paradox::ParamSet |
list())
Configuration parameter set or dynamic parameter-set source list.- packages
(
character())
Required packages.- label
(
character(1))
Label.- man
(
character(1))
Help page reference.- additional_phash_input
(
character())
Additional private/public fields used for persistent hashing.- context
(ALContext)
Active-learning proposal context.- n
(
integer(1))
Maximum number of points to propose.
Fields
label(
character(1)) Label.man(
character(1)) Help page reference.packages(
character()) Required packages.