dynparam: Creating meta-information for parameters

In dynverse/dynparam: Creating Meta-Information for Parameters

Description

Provides tools for describing parameters of algorithms in an abstract way. Description can include an id, a description, a domain (range or list of values), and a default value. 'dynparam' can also convert parameter sets to a 'ParamHelpers' format, in order to be able to use 'dynparam' in conjunction with 'mlrMBO'.

Parameter set

• Create a new parameter_set() by adding several parameters to it

• as_paramhelper(): Convert it to a ParamHelpers object

• sip(): Sample a parameter set

Parameters

These functions help you provide a meta description of parameters.

Implemented are the following functions:

• character_parameter(), integer_parameter(), logical_parameter(), numeric_parameter(): Creating parameters with basic R data types.

• integer_range_parameter(), numeric_range_parameter(): Create a discrete or continuous range parameter.

• subset_parameter(): A parameter containing a subset of a set of values.

See ?parameter for a list of helper functions converting parameters from and to other formats.

Distributions

These distributions allow to define prior distributions for numeric and integer parameters.

Implemented are the following distributions:

• uniform_distribution()

• expuniform_distribution()

• normal_distribution()

See ?distribution for a list of helper functions converting parameters from and to other formats.

Advanced topics

• distribution(): Creating a custom distribution

dynverse/dynparam documentation built on May 29, 2019, 8 a.m.