Skip to contents

This class facilitates interactive visual analysis of parameter scan results.

See also

Public fields

analyzed

List, as output by analyze_parameter_scan().

params

Vector of names of scanned parameters.

metrics

Vector of names of model performance metrics to use.

n_params

Number of scanned parameters.

n_metrics

Number of performance metrics to apply.

res

data.frame holding parameter scan results. It contains n_params + n_metrics + 1 columns: one column for each scanned parameter, one for each employed metric and an additional column (name n) to give each parameter combination (i.e. each row) an identifying number.

n_combinations

Number of rowns in res.

sorted

List containing copies of res which are each sorted by a different performance metric. List keys are the values in self$metrics.

selection

Vector of integers corresponding to the ID number of combinations (column n in self$res) that is to be highlighted.

Methods


Method new()

Construct and set up a PscanPlotter instance.

Usage

PscanPlotter$new(analyzed, variable = "dBM")

Arguments

analyzed

List; Output of analyze_parameter_scan().

variable

Str; Name of variable in analyzed to visualize. Can be changed later with set_variable(). Allowed values are the keys in analyzed except for params and metrics.


Method set_variable()

Choose which variable to visualize.

Usage

PscanPlotter$set_variable(variable)

Arguments

variable

Chosen variable name. One of "dBM", "cBM", "cBM_end"


Method analyze()

Enter analysis loop.

This plots the analysis results and enters a simple command-line interface through which more insights can be gathered. Particularly, it allows highlighting specific parameter combinations, either by their index number or by selecting the best performers according to a given metric.

Usage

PscanPlotter$analyze()


Method plot()

Plot parameter scan results.

For every combination of scanned parameter and metric, a subplot is generated in which the parameter values are plotted against performance score in that metric for every parameter combination.

The result of this is static. Use PscanPlotter$analyze() for an interactive version.

Usage

PscanPlotter$plot()


Method print_info()

Print information on selected parameter combinations.

The parameter values and performance scores of all combinations referred to by the integers in selection are printed to console.

Usage

PscanPlotter$print_info(selection)

Arguments

selection

Vector of integers representing IDs of parameter combinations (i.e. column n in self$res).


Method clone()

The objects of this class are cloneable with this method.

Usage

PscanPlotter$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.